Flavor, Fragrance, and Odor Analysis

ISBN: 0-8247-0627-7
This book is printed on acid-free paper.
Headquarters
Marcel Dekker, Inc.
270 Madison Avenue, New York, NY 10016
tel: 212-696-9000; fax: 212-685-4540
Eastern Hemisphere Distribution

Marcel Dekker AG
Hutgasse 4, Postfach 812, CH-4001 Basel, Switzerland
tel: 41-61-261-8482; fax: 41-61-261-8896
World Wide Web

http:/ /www.dekker.com
The publisher offers discounts on this book when ordered in bulk quantities. For more
information, write to Special Sales/Professional Marketing at the headquarters address
above.
Copyright 2002 by Marcel Dekker, Inc. All Rights Reserved.
Neither this book nor any part may be reproduced or transmitted in any form or by any
means, electronic or mechanical, including photocopying, microfilming, and recording,
or by any information storage and retrieval system, without permission in writing from
the publisher.
Current printing (last digit):

10 9 8 7 6 5 4 3 2 1
PRINTED IN THE UNITED STATES OF AMERICA

To Deb
Equipped with his five senses, man explores the universe around him
and calls the adventure science.
Edwin P. Hubble
Preface
This book focuses on recent sample preparation techniques for isolating and con-
centrating flavor and odor chemicals from various types of foods, beverages, and
consumer products prior to gas chromatography (GC)/mass spectrometry (MS)
analysis. No single sample preparation technique is appropriate for every type
of analyte or matrix. We show the advantages, disadvantages, and biases of the
most common analytical techniques for flavor, fragrance, and odor analysis. The
intent of this book is to help chemists working with flavor, fragrance, and odor
problems to select the most appropriate techniques for studying specific applica-
tions. This text explores the application potential of various analytical techniques,
including numerous practical examples and tips that explain how state-of-the-art
techniques can be used to resolve important flavor, fragrance, and odor issues
facing chemists in the food and beverage and consumer product industries.
The book can be categorized into three parts: sample preparation and instru-
mentation techniques, application examples, and olfactometry. The final two
chapters discuss MS-based electronic nose applications (Chapter 13) and the
chemical structures of flavor and off-flavor chemicals in various types of foods
(Chapter 14).
This book follows up Techniques for Analyzing Food Aroma, and a few
chapters that discuss standard GC/MS sample preparation techniques have been
taken from that work. Some of the analytical techniques discussed in Techniques
for Analyzing Food Aroma were well established, while techniques such as solid-
phase microextraction (SPME) and electronic-nose applications were emerging
technologies. In recent years, more and more researchers have discoverd the nu-
merous advantages of SPME, and its popularity and use in extracting and concen-
trating flavor/odor-contributing analytes have skyrocketed. Several chapters in
this book emphasize SPME techniques.
v
vi Preface
For various reasons, reliable electronic-nose applications have been slower

to develop. Chapter 13 discusses the benefits of MS as a potential e-nose sensor.
This book also discusses the value of time-of-flight MS to the study of flavors
and odors. Incorporating the human sense of smell with potent analytical systems
is invaluable in problem solving. Just as sample preparation procedures and ana-
lytical instrumentation have continued to evolve and improve, so have olfactome-
try techniques. Chapters 11 and 12 cover various olfactometry techniques, includ-
ing a new, easier-to-implement method called SNIF.
I commend the contributing authors for their dedication, persistence, and
cooperation in completing their chapters in a timely manner. Unquestionably, the
information they have provided in this book, as well as in past publications, will
contribute greatly to the advancement of flavor and fragrance research. I would
also like to acknowledge my wife, Deborah, for her review of the chapters, her
patience, and her continuous support and words of encouragement.
Ray Marsili
Contents
Preface v
Contributors ix
1. Solvent Extraction and Distillation Techniques 1

Thomas Parliment
2. Analysis of Food Volatiles Using Headspace-Gas

Chromatographic Techniques 25
Thomas P. Wampler
3. The Analysis of Food Volatiles Using Direct Thermal

Desorption 55
Casey C. Grimm, Steven W. Lloyd, James A. Miller, and Arthur
M. Spanier
4. Solid-Phase Microextraction for the Analysis of Aromas and

Flavors 75
Alan D. Harmon
5. The Advantages of GC-TOFMS for Flavor and Fragrance

Analysis 107
John F. Holland and Ben D. Gardner
vii
viii Contents
6. Modern Methods for Isolating and Quantifying Volatile Flavor

and Fragrance Compounds 139
Peter Werkhoff, Stefan Brennecke, Wilfried Bretschneider, and
Heinz-Jurgen Bertram
7. SPME Comparison Studies and What They Reveal 205

Ray Marsili
8. Analysis of Volatile Compounds in the Headspace of Rice

Using SPME/GC/MS 229
Casey C. Grimm, Elaine T. Champagne, and Kenichi Ohtsubo
9. Headspace Techniques for the Reconstitution of Flower Scents

and Identification of New Aroma Chemicals 249
Thomas McGee and Kenneth L. Purzycki
10. SPME Applications in Consumer Products 277

Richard Payne, Allen E. Puchalski, and John Labows
11. Gas ChromatographyOlfactometry in Food Aroma Analysis 297

Imre Blank
12. Quantitative Use of Gas ChromatographyOlfactometry: The

GC-SNIF Method 333
Alain Chaintreau
13. Combining Mass Spectrometry and Multivariate Analysis to

Make a Reliable and Versatile Electronic Nose 349
Ray Marsili
14. Character Impact Compounds: Flavors and Off-Flavors in Foods 375

Robert J. McGorrin
Abbreviations 415
Index 419
Contributors
Heinz-Jurgen Bertram Haarmann & Reimer GmbH, Holzminden, Germany
Imre Blank Nestle Research Center, Lausanne, Switzerland
Stefan Brennecke Haarmann & Reimer GmbH, Holzminden, Germany
Wilfried Bretschneider Haarmann & Reimer GmbH, Holzminden, Germany
Alain Chaintreau Firmenich S.A., Geneva, Switzerland
Elaine T. Champagne Southern Regional Research Center, Agricultural Re-

search Service, U.S. Department of Agriculture, New Orleans, Louisiana
Ben D. Gardner Orion Associates Science & Engineering, Okemos, Michigan
Casey C. Grimm Southern Regional Research Center, Agricultural Research

Service, U.S. Department of Agriculture, New Orleans, Louisiana
Alan D. Harmon Research and Development, McCormick & Co., Inc., Hunt
Valley, Maryland
John F. Holland Department of Biochemistry, Michigan State University, East

Lansing, Michigan
ix
x Contributors
John Labows Colgate-Palmolive Company, Piscataway, New Jersey
Steven W. Lloyd Southern Regional Research Center, Agricultural Research

Ray Marsili Dean Foods Company, Rockford, Illinois
Thomas McGee Global Technology and Innovation, Givaudan Fragrances

Corp., Teaneck, New Jersey
Robert J. McGorrin Department of Food Science and Technology, Oregon

State University, Corvallis, Oregon
James A. Miller Southern Regional Research Center, Agricultural Research

Kenichi Ohtsubo National Food Research Institute, Ministry of Agriculture,

Forestry and Fisheries, Ibaraki-Ken, Japan
Thomas Parliment Parliment Consulting, New City, New York
Richard Payne Colgate-Palmolive Company, Piscataway, New Jersey
Allen E. Puchalski Colgate-Palmolive Company, Piscataway, New Jersey
Kenneth L. Purzycki Givaudan Fragrances Corp., Teaneck, New Jersey
Arthur M. Spanier Beltsville Agricultural Research Center, Agricultural Re-

search Service, U.S. Department of Agriculture, Beltsville, Maryland
Thomas P. Wampler CDS Analytical, Inc., Oxford, Pennsylvania
Peter Werkhoff Haarmann & Reimer GmbH, Holzminden, Germany

1
Solvent Extraction and Distillation
Techniques
Thomas Parliment
Parliment Consulting, New City, New York
I. INTRODUCTION
The purpose of this chapter is to review techniques that have been published
in the technical literature and developed in our laboratory for the isolation and
concentration of samples prior to analysis by gas chromatography. It is our goal
to emphasize those techniques that are easy to employ, require minimal equip-
ment, and produce reproducible, meaningful results. In a number of cases, exam-
ples of the results will be presented.
As has been described previously (1), sample preparation is complicated
by a number of factors:
1. Concentration Level: Aromatics levels are generally low, typically in
the ppm, ppb, or ppt range. Thus, it is necessary not only to isolate
the components but also to concentrate them by several orders of mag-
nitude.
2. Matrix: The volatiles are frequently intracellular and must be liberated
by disruption. The sample frequently contains nonvolatile components
such as lipids, proteins, or carbohydrates, which complicates the isola-
tion process. These components may create problems of foaming and
emulsification during isolation procedures and will create artifacts if
injected into a hot gas chromatography injector port.
3. Complexities of Aromas: The aromatic composition of foods are fre-
quently very complex. For example, coffee currently has almost 800
identified components, as shown in Table 1. Complicating the picture
1
2 Parliment
TABLE 1 Classes of Aroma Compounds in Coffee
Number of
Chemical class compounds
Hydrocarbons 74
Alcohols 20
Aldehydes 30
Ketones 73
Acids 25
Esters 31
Lactones 3
Phenols (and ethers) 48
Furans 127
Thiophenes 26
Pyrroles 71
Oxazoles 35
Thiazoles 27
Pyridines 19
Pyrazines 86
Amines and miscellaneous nitrogen compounds 32
Sulfur compounds 47
Miscellaneous 17
Total 791
Source: Ref. 2.
is the fact that the classes of compounds present cover the range of
polarities, solubilities, and pHs.
4. Variation of Volatility: The components possess boiling points ranging
from well below room temperature to those that are solids, such as
vanillin (mp 81C).
5. Instability: Many components in an aroma are unstable and may be
oxidized by air or degraded by heat or extremes of pH.
Regardless of which sample preparation technique is employed, it is criti-
cally important to assess the organoleptic quality of the isolate. No single tech-
nique will prove optimal for every sample, and evaluations should be made to
ensure that decomposition and loss of desired components do not occur. A very
significant paper published by Jenings et al. (3) compared various sample prepara-
tion techniques, including porous polymer trapping and distillation-extraction.
Their conclusion was that no isolation technique produced results that duplicated
the original neat sample, but that distillation-extraction most nearly agreed
(Fig. 1).
Solvent Extraction and Distillation Techniques 3
FIGURE 1 Relative integrator response for various sample preparation techniques. (From
Ref. 3.)
4 Parliment
This is particularly important since current flavor research seems to be less

directed to identification for the sake of adding to the numbers of the compounds
in the knowledge base, and more to alternative reasons. At the present time it
appears one purpose is characterization of components of organoleptic impor-
tance. Three techniques for gas chromatographic individual component assess-
ment are in vogue: aroma extraction dilution analysis (AEDA), calculation of
odor units, and CharmAnalysis (see Chapter 12). Another purpose of flavor re-
search is to analyze products and to perform flavor stability studies.
At the present time, the two most common procedures reported in the litera-
ture for the isolation of the aromatics are headspace methods and extraction. The
former will be covered in the next chapter. The purpose of this chapter is to
review techniques for isolating and concentrating aromatics, which include vari-
ous distillation and extraction procedures.
A number of references exist on the topic of flavor isolation, and these
provide a different perspective on the topic (48). To quote Schreier (9): It
must be emphasized that sample preparation is the most critical step in the entire
analytical process of the investigation of volatiles.
II. DIRECT INJECTION OF THE SAMPLE

A. Essential Oils
Direct injection is by far the most convenient technique and works particularly
well for essential oils. The sample may have to be diluted with a solvent to obtain
response within the limits of the detector.
B. Aqueous Samples
When concentrated aqueous samples are available, direct injection techniques
can be employed. In industry, aqueous materials are frequently available from
industrial operations. Examples of this would be condensates from coffee grind-
ers, vapors from chocolate conching operations, and aqueous materials from cit-
rus juice concentrators.
The aqueous phase may be injected if the sample is sufficiently concen-
trated. A number of problems may be encountered under these circumstances.
When water is converted to steam, the volume increases dramatically; 1 l of
water becomes more than 1000 l of steam. This is larger than the injector volume
of many current gas chromatographs, and the steam may degrade the performance
of the system. Polar gas chromatography liquid phases such as Carbowax and
PEG will degrade in the presence of steam unless they are bonded to the column.
If the aqueous sample contains dissolved solutes such as carbohydrates or
proteins, additional problems will arise when the sample is injected. The nonvola-
tiles may decompose, leaving a nonvolatile residue in the injector and at the head
of the column. Many researchers use a guard column of deactivated fused silica
tubing between the injector and the analytical column. The guard column can be
replaced periodically when it becomes contaminated. The tubing contains no liq-
uid phase, thus it does not affect separation or retention time. The guard column
can be connected to the analytical column with various types of press-tight con-
nectors (10).
If the aqueous phase is too dilute, concentration techniques as described
in the next section may be employed.
III. DIRECT SOLVENT EXTRACTION OF AQUEOUS

SAMPLES
Aqueous samples are available from a number of sources. Industrial plant opera-
tions may yield such products. Carbonated beverages, fruit juices, and caffeinated
beverages can often be extracted directly. Fruits and vegetables can be homoge-
nized with water, treated with a pectinase enzyme to destroy the pectins, and
filtered through a bed of diatomaceous earth to remove particulates.
A. Extraction
When relatively large amounts of aqueous samples are available, then separatory
funnels or commercial liquid-liquid extractors may be employed. A large number
of solvents have been summarized by Weurman (4) and reviewed by Teranishi
et al. (5).
The solvents most commonly used today are diethyl ether, diethyl ether/
pentane mixtures, hydrocarbons, Freons, and methylene chloride. The latter two
have the advantage of being nonflammable. Solvent selection is an important
factor to consider, and the current status has been summarized by Leahy and
Reineccius (11). In general, the following suggestions can be made. Nonpolar
solvents such as Freons and hydrocarbons should be used when the sample con-
tains alcohol. Diethyl ether and methylene chloride are good general purpose
solvents. Ether can form explosive peroxides, and for that reason contains inhibi-
tors (e.g., BHT), which will show up in gas chromatography/mass spectroscopy
(GC/MS) analysis. We find that methylene chloride is a satisfactory general pur-
pose solvent, particularly for flavor compounds with an enolone structure (e.g.,
Maltol and Furaneol). It is somewhat toxic and is an animal carcinogen. To aid
in extraction, sodium chloride may be added to the aqueous phase to salt out the
organics when low-density solvents are employed.
If the sample contains any particulates, it should be filtered. A convenient
way to filter samples is through a syringe filter (e.g., Gelman Sciences, Ann
6 Parliment
Arbor, Mich.) of the type recommended for HPLC sample preparation. These
filters have a pore size of 0.45 m and are solvent resistant. Microtypes with low
solvent hold-up are available.
Figure 2 shows the total ion chromatogram of a coffee extract. In this case
a decaffeinated roast and ground coffee was brewed in a commercial system.
The brew was filtered through a Gelman 0.45 m GHP Acrodisc to remove partic-
ulates, and the aqueous phase was extracted with methylene chloride. A highly
complex chromatogram is evident. The large peak eluting at 25 minutes is caf-
feine.
Continuous extractors have been described in the literature for solvents
more dense and less dense than water (e.g., Ref. 4) and are available commer-
cially (e.g., ACE Glass, Vineland, NJ; Supelco, Inc., Bellefonte, Pa) for $200
600 (Fig. 3). These are a pleasure to use (providing there is no solvent loss and
that emulsions dont occur) since they will operate relatively unattended. They
are normally operated for 24 hours, but may be operated overnight.
Liquid carbon dioxide was recommended as an extracton solvent as early
as 1970 (12). It has the advantages of being nontoxic and inexpensive. Liquid
carbon dioxide is reported to have solvent properties similar to diethyl ether (12)
and to be particularly selective for esters, aldehydes, ketones, and alcohols. If
water is present, it will be removed also.
A commercial liquid carbon dioxide Soxhlet extractor is commercially
available (J&W Scientific, Folsom, CA). The vessel holds a sample of 2.5 g.
FIGURE 2 Total ion chromatogram (TIC) of brewed R&G coffee extracted with methy-
lene chloride.
FIGURE 3 Liquid/liquid extractor concentrator apparatus. (Courtesy Supelco, Inc.,

Bellefonte, PA.)
This apparatus seems to have achieved only limited use, perhaps because of its
cost ($1500 plus accessories) and limited sample size. Moyler (13) discussed a
commercial liquid carbon dioxide system and reported such extracts to be more
concentrated than the steam distillates or solvent extracts. More important, he
reported that the character was finer.
Supercritical carbon dioxide has been employed recently as an extraction
solvent. When using supercritical carbon dioxide, it is necessary to balance tem-
perature, pressure, and flow rate, which requires complex instrumentation. Sev-
8 Parliment
eral instrument vendors produce supercritical fluid extractors in the price range
of $25,00090,000. Again, sample capacity is relatively limited.
B. Emulsions
Emulsions can be a problem, particularly if nonvolatile solutes are present. To
prevent emulsions, the following methods can be employed:
Use gentle shaking.
Filter the sample if particulates are present.
Keep the system cool.
Be patient.
Adjust the pH of the aqueous phase.
The latter technique is particularly effective if organic acid, basic, or amphoteric
compounds are present. If emulsions occur, centrifugation may be employed (but
only for nonflammable solvents).
C. Concentration
The final step is concentration of the solvent. We usually dry the solvent over
sodium sulfate or magnesium sulfate and then carefully concentrate it on a steam
bath using a Vigreux column. A convenient method to concentrate large volumes
of solvent is by use of a Kuderna-Danish Evaporative Concentrator, which is
available in both macro (up to 1000 ml) and micro (14 ml) capacities for less
than $100.
D. Impurities
High-boiling impurities both in solvent and sample will also be concentrated
along with the desired analytes. Thus, solvent blanks should be prepared. If the
sample was a direct extract, the solvent will contain nonvolatile components such
as natural and Maillard pigments, lipids, alkaloids, etc. These may crystallize or
precipitate on concentration and will leave a residue in the injector of the gas
chromatograph.
For additional suggestions on extracting aqueous samples, see Secs. IV
and V.
IV. STEAM DISTILLATION OF SAMPLES FOLLOWED

BY SOLVENT EXTRACTION
One of the most common sample-preparation techniques employed today in-

volves steam distillation followed by solvent extraction. The primary advantage
is that the distillation step separates the volatiles from the nonvolatiles. Other
reasons for this include simplicity of operation, no need for complex apparatus,
reproducibility, rapidity, and the range of samples that can be handled. Steam
distillation works best for compounds that are slightly volatile and water insolu-
ble. In addition, compounds with boiling points of less than 100C will also pass
over.
A. Direct Distillation
The sample is normally placed in a round-bottom flask and dispersed in water.
The aqueous slurry can be heated directly (with continuous stirring) to carry over
the steam-distillable components. Problems can be encountered due to scorching
of the sample if too much heat is applied, and in addition bumping may occur
when the sample contains particulates. Stirring may prevent these problems.
Foaming is another potential problem. Many food products contain surface-active
agents and will foam during distillation; addition of antifoams (e.g., DC polydi-
methyl siloxanes) may prevent this problem, but these silicones usually end up
in the distillate, as evidenced by GC-MS peaks at m/z 73, 147, 207, 221, 281,
and 341.
B. Indirect Steam Distillation

Indirect steam distillation has many advantages over the direct technique. It is
more rapid, and less decomposition of the sample occurs because the sample is
not heated directly. The steam may be generated in an external electrically heated
steam generator or in a round-bottom flask heated by a mantle. It is even possible
to use laboratory house steam, in which case the steam must be passed through
a trap that allows removal of condensate and any particulates that may come out
of the line. It is imperative that blank samples be run, since house steam may
be highly contaminated. Even so, this technique has the great advantage of being
rapid and easy. The steam and volatiles are usually condensed in a series of traps
cooled with a succession of coolants ranging from ice water to dry ice/acetone
or methanol.
C. Vacuum Steam Distillation

If sample decomposition remains a concern, then the steam distillation may be
operated under vacuum. In this case inert gas should be bled into the system to
aid in agitation. A number of cooled traps should be in line to protect the pump
from water vapor and the sample from pump oil vapors. Another simple method
to generate a condensate under vacuum is by use of a rotary evaporator. Bumping
10 Parliment
is normally not a problem in this case. The higher-boiling components do not

distill as efficiently as they do under atmospheric pressure.
Once the vapors have been condensed, it remains to extract the sample,
which is normally very dilute. Techniques described in Sec. III may be em-
ployed. In addition, there are two semi-micro extraction techniques that have
value.
D. Extraction
Use of the Mixxor has been described by Parliment (14) and its utility described
in a number of publications (15,16). Such a device is shown in Fig. 4. These
extractors are available with sample volumes ranging from 2 ml to 100 ml. The
10-ml capacity extractor is a particularly convenient capacity for flavor research.
Briefly, approximately 8 ml of aqueous condensate is placed in receiver B and
saturated with sodium chloride. The whole assembly is cooled and then a quantity
of diethyl ether (typically 0.50.8 ml) is added. The ether may contain an internal
standard. The system is extracted by moving chamber A up and down a number
of times. After phase separation occurs, the solvent D is forced into an axial
chamber C, where it can be removed with a syringe for analysis. Percent recover-
FIGURE 4 Mixxor apparatus for the extraction of aqueous samples. (From Ref. 14.)
ies for a series of ethyl esters from an aqueous solution were essentially quantita-
tive even at the sub-ppm level. These extractors are currently available from
Sigma-Aldrich, Milwaukee, WI 53223.
A less sophisticated alternative exists. The sample may be placed in a
screw-capped centrifuge tube and a small amount of dense solvent added. After
exhaustive shaking, the tube can be centrifuged to break the emulsion and sepa-
rate the layers. The organic phase can be sampled from the bottom of the tube
with a syringe. Methylene chloride works well in this application.
Figure 5 compares the total ion chromatograms of two samples. Roasted
and ground coffee was indirectly steam distilled at atmospheric pressure and a
condensate collected. The upper curve in the figure represents the ethereal con-
centrate prepared via the Mixxor technique; the lower curve is the methylene
chloride extract. Pattern differences are apparent. The largest peak in the ethereal
extract (Rt 6.0) is furfuryl alcohol; the largest peak in the lower curve (Rt
8.5) is 5-methyl furfural.
E. Manipulation of the Aqueous Phase

Adjustment of the pH of the aqueous phase before extraction may accomplish
two goals. First, emulsions may be broken, permitting phase separation to take
place rapidly. Second, class separation will take place, which may simplify the
gas chromatographic pattern. This is less necessary today because contemporary
gas chromatography columns have high resolving power; frequently, however,
small peaks are concealed under larger ones, and the smaller ones may be re-
vealed for organoleptic evaluation or identification.
This chemical manipulation of the aqueous phase can be carried even fur-
ther. Many food aromatics contain carbonyl compounds. By adding sodium bisul-
fite to the aqueous phase, it is possible to selectively remove the aldehydes and
the methyl ketones by forming their water-soluble bisulfate addition complexes.
Thus, this analysis produces a carbonyl-free sample.
Figure 6 shows an example of such a manipulation. In this case the upper
curve is the ethereal extract of a steam distillate of coffee at pH 3.6. The larger
asymmetrical peaks at Rt 5.8 and 6.8 represent fatty acids. These are eliminated
at pH 10.0 (middle curve). The large peak at Rt 3.8 in the latter is pyridine,
a decomposition product of trigonilline. The lower curve in this figure is the
material that remains after bisulfite extraction. It is immediately apparent that
many of the lower boiling components of coffee are carbonyl in nature. For exam-
ple, the peak at Rt 5.5 is furfural. In this manner it is possible to simplify the
gas chromatographic pattern.
If the aqueous phase is limited in quantity, the analyst can perform an
interesting set of sequential experiments. The sample is placed in the Mixxor
12 Parliment
FIGURE 5 Comparison of chromatograms of etheral (upper) and methylene chloride

(lower) extract of R&G coffee.
Chamber B, the pH adjusted to about 3 with acid, and the sample extracted with
diethyl ether. Sufficient sample is removed for gas chromatographic analysis,
e.g., 1 l. The aqueous phase is made alkaline and the sample reextracted with
the same diethyl ether and gas chromatographic analysis repeated. Finally, the
sample is made neutral and saturated with sodium bisulfite and reextracted. The
ethereal phase is reanalyzed. In this case three different analyses can be made
from the same sample in a short period of time and subjected to GC-MS and
organoleptic analysis.
FIGURE 6 Comparison of chromatograms of R&G coffee extracted at pH 3.6 (upper

curve), pH 10.0 (middle curve), and with bisulfite (lower curve).
14 Parliment
V. SIMULTANEOUS STEAM DISTILLATION/EXTRACTION
One of the most popular and valuable techniques in the flavor analysis field is
the simultaneous steam distillation/extraction (SDE) apparatus first described by
Likens and Nickerson (17). The apparatus provides for the simultaneous conden-
sation of the steam distillate and an immiscible organic solvent. Both liquids are
continuously recycled, and thus the steam distillable-solvent soluble compounds
are transferred from the aqueous phase to the solvent. The advantages of this
system include the following:
FIGURE 7 Various modifications to SDE apparatus. (From Ref. 9.)

1. A single operation removes the volatile aromas and concentrates them.

2. A small volume of solvent is required, reducing problems of artifact
buildup as solvents are concentrated.
3. Recoveries of aroma compounds are generally high.
4. The system may be operated under reduced pressure to reduce thermal
decomposition.
A number of refinements have been made to the basic apparatus, some of which
are shown in Fig. 7ad (9).
Typically the sample flask has a 500 ml to 5 liter capacity and contains
the sample dissolved or dispersed in water so that the flask is less than half filled.
Agitation is advisable if suspended materials are present to prevent bumping. As
with all distillations, the pH of the sample should be recorded (and adjusted if
necessary) prior to distillation. Heat may be supplied by a heating mantle or
(better if solids are present) a heated oil bath with stirrer. The solvent is normally
contained in a pear-shaped flask of 1050 ml capacity. Many solvents have been
employed. In one model system study, Schultz et al. (18) compared various sol-
vents as the extractant. They reported that hexane was an excellent solvent except
for lower-boiling water-soluble compounds, where diethyl ether was considerably
better. Use of methylene chloride has been recommended in a modified Likens-
Nickerson extractor (19). Currently, most researchers appear to be using pentane-
diethyl ether mixtures.
Regardless of which solvents are used, boiling chips should be added to
both flasks to ensure smooth boiling. The distillation is generally performed for
13 hours. After the distillation is completed, the system is cooled and the solvent
from the central extracting U tube is combined with that of the solvent flask. The
solvent is dried over an agent such as sodium sulfate and concentrated by slow
distillation.
An impressive example of the use of a Likens-Nickerson extractor is shown
in Fig. 8. This figure shows the gas chromatogram of a green and a roasted
Kenyan coffee and shows how aromatic compounds are generated in the roasting
process (W. Holscher, personal communication).
Vacuum versions of the SDE system have been described. These have
the advantage of reducing the thermal decomposition of the analyte. Leahy and
Reineccius report (11) that vacuum operation had a slightly negative effect upon
recovery compared to atmospheric operation. Our experience is that operation
under vacuum is quite complex since one must balance the boiling of two flasks,
keep the solvent from evaporating, and hold the pressure constant.
Table 2 presents results of a series of experiments wherein typical flavor
compounds in a model mixture were isolated by various SDE techniques. In
general, ether is a better solvent than hydrocarbons, and atmospheric pressure
better than reduced pressure.
16 Parliment
FIGURE 8 Chromatographic comparison of green and roasted coffee. (W. Holscher, per-
sonal communication.)
VI. DIRECT SOLVENT EXTRACTION OF SOLID SAMPLES
An entirely different process of sample-preparation technique involves direct sol-

vent extraction, which is a very simple and convenient technique. Probably the
easiest way to do such an extraction is with a Soxhlet extractor. A dried sample
such as a spice, chocolate nib, R&G coffee, or a grain can be ground finely and
placed in a Soxhlet thimble and extracted with an organic solvent. Either diethyl
ether or methylene chloride may be used in such a system. After a number of
cycles, the solvent can be combined and concentrated. Nonvolatile organic mate-
rials such as lipids, alkaloids such as caffeine and theobromine, and pigments
will also be concentrated. The sample may be analyzed directly (with trepidation)
or it may be treated as described in the section below, after removal of the solvent.
If the sample contained large amounts of lipids (e.g., coffee, chocolate), then the
volatiles may be removed by subsequent steam distillation or by a high vacuum
stripping technique as described in Sec. VII.
Figure 9 is the GC-MS of a roast and ground coffee sample, which was
moistened with water and extracted with methylene chloride in a Soxhlet extrac-
tor. The large component eluting at 26 minutes is caffeine.
Solvent Extraction and Distillation Techniques
TABLE 2 Recovery of Components by SDE from the Model Mixture at a Concentration of 165 ppm (w/v) for Each Compound
(Recovery as Percentage of Initial Amount)
Times of SDE 1 hr 4 hr
Pressure Atmospheric pressure 100 mm Atm
a
Vol. of solvent 125 ml 10 ml 125 ml 125 ml
Solvent Hexane Pentane Ether Hexane Hexane Hexane
pH 3.4 5.0 6.5 7.8 5.0b 5.0 5.0 5.0 5.0 5.0
Ethyl acetate 0 0 0 0 0 59 89 19 0 0
Ethyl butyrate 98 99 99 91 99 101 97 84 100 98
Ethyl hexanoate 100 101 101 95 101 102 99 97 103 99
Ethyl octanoate 99 99 100 95 100 102 100 99 100 99
Ethyl 3-hydroxy- 41 41 41 19 42 44 49 30 6 90
hexanoate
Ethanol 0 0 0 0 0 0 58 0 0 0
1-Hexanol 101 101 103 98 100 102 100 96 98 100
Linalool 73 99 100 96 99 99 97 97 99 98
Octanol 102 102 103 98 102 103 101 99 103 101
Citronellal 59 78 98 94 81 81 79 77 95 80
Carvone 98 97 98 95 98 99 97 97 92 99
a
For this run, additional hexane (13 ml) was added through the vent to fill the overflow arm before the distillation was started, and the extract was not
concentrated after SDE.
b
1.0 ml of glacial acetic acid, titrated in solution to pH 5 with sodium hydroxide, was also present in this run in addition to the usual citrate buffer at
0.05 M.
Source: Ref. 18.
17
18 Parliment
FIGURE 9 TIC of a roast and ground coffee sample moistened with water and extracted
with methylene chloride.
VII. HIGH VACUUM DISTILLATION OF LIPIDS
A number of the procedures described in Sec. VI will yield a material that is

primarily lipid in nature. In addition, many samples available to the researcher
are themselves lipids. A few materials that one may encounter are coffee oil,
vegetable and nut oils, cocoa butter, lard, butter oil, lipids used for deep fat frying,
and lipids used as the solvent for Maillard reaction systems. Such materials can
be a relatively rich source of aromatic compounds because aroma compounds
are typically lipid soluble. A number of procedures can be used to prepare a
sample. In this section we will cover three useful ones.
A. Steam Distillation
The lipid material may be steam distilled at atmospheric pressure or under vac-
uum, as was described in Sec. IV, and subsequently subjected to solvent extrac-
tion. Alternatively, a modified Likens-Nickerson extractor has been described
(19), which permits the introduction of steam into the system. Recoveries of
model compounds from lipid systems were not as satisfactory as for aqueous
samples.
FIGURE 10 Falling film molecular still for the removal of volatiles from lipids. (From
Ref. 20.)
B. High Vacuum Distillation

When large amounts of lipid materials are present, the sample may be subjected
to a falling film molecular still. The apparatus utilizes the principle of vaporiza-
tion of the flavor from a heated thin film of the oil under high vacuum. One such
apparatus is shown in Fig. 10 (20). Several hundred milliliters of oil are placed in
vessel A and slowly passed through the foaming chamber into the heated bellows
chamber. The distillate is collected in a series of traps cooled with liquid nitrogen.
The oil may be recycled. Another series of apparatus described by Chang et al.
at Rutgers (21) has accomplished similar goals. This type of apparatus generally
falls into the same category of equipment as that used to deodorize lipids.
C. Short Path Distillation

One version of the apparatus is shown in Fig. 11a. The nonvolatile material is
placed in the flask. The flask is heated while stirring the sample and a high vac-
20 Parliment
FIGURE 11 Apparatus for the removal of aromatics from lipids. (a from Ref. 22; b from
Ref. 23.)
uum is applied. The inner condenser is cooled with liquid nitrogen or dry ice-
solvent (22). We have found this apparatus very useful for separating the volatile
aromatics from nonvolatile residues (i.e., lipids) such as those generated in Sec-
tion VI. In that case the sample size may be only a few grams or less, and a smaller
version of the short path distillation apparatus is appropriate. This apparatus can
be easily fabricated by a glassblower.
An example of the application of such an apparatus is shown in Fig. 12.
The sample was produced by high vacuum distillation of 10 g of coffee oil ex-
pelled from roast and ground coffee. The volatiles were condensed with liquid
nitrogen and subsequently washed off the cold finger with methylene chloride.
Figure 12 shows the total ion chromatogram of the sample. The large peak eluting
at 25 minutes is caffeine.
Nawar (23) has commented that the apparatus shown in Fig. 11a may pres-
ent problems if the sample contains water. He suggested the apparatus shown in
Fig. 11b. Vacuum is applied at point A, and vessel L is filled with liquid nitrogen
during the 1-hour distillation period. At the end of the distillation, the cold trap
is disconnected and the coolant is discarded. The condenser is inverted, the ice
melted, and condensed volatiles and water extracted with a solvent. He reported
greater than 80% recovery of high-boiling hydrocarbons in a model system study.
VIII. CO-DISTILLATION OF SAMPLE WITH SOLVENT
A new technique has been suggested by a group of Russian workers (24). They
compared three methods of isolation, namely, distillation-extraction and two
methods based on co-distillation of sample from solvent-water mixtures. In the
FIGURE 12 TIC of volatiles from roast and ground coffee oil, distilled in apparatus shown
in Fig. 11a.
co-distillation technique, a solvent such as diethyl ether, pentane, or methylene

chloride is dispersed in the sample and the sample is distilled rapidly (at, e.g.,
200C) until all the solvent and a small amount of water have passed over. The
sample is analyzed by gas chromatography. Their co-distillation technique
(at atmospheric pressure) compared favorably with the Likens-Nickerson tech-
nique. They analyzed three samples: a model system, a meat sample, and a fish
sample.
The chromatogram of an R&G coffee that was dispersed in water and co-
distilled with solvent in our laboratory is presented in Fig. 13. The curve is the
total ion chromatogram of the sample, which has large caffeine peak eluting at
25 minutes.
The advantages of co-distillation are that isolates are generated without a
boiled note, the process is efficient and reproducible, and it takes only 1520
minutes for a distillation.
IX. SUMMARY
Over the years numerous procedures have been proposed for the isolation and
identification of aromatic compounds. Because of the variation of sample types
encountered, no single technique will always suffice. One must always be aware
that none of these techniques will produce an isolate that quantitatively represents
the composition of the starting material.
22 Parliment
FIGURE 13 TIC of a roast and ground coffee sample co-distilled with methylene chloride.
This chapter reviewed techniques that involve distillation and extraction

procedures. These have the advantage of being simple and rapid, and they do
not require a complex apparatus. For typical food products some version of the
Likens-Nickerson distillation apparatus is probably the technique of choice; for
lipid materials, some high-vacuum distillation procedure is worth investigating
initially.
REFERENCES
1. T. H. Parliment, Sample Preparation Techniques for Gas-Liquid Chromatographic

Analysis of Biologically Derived Aromas, in Biogeneration of Aromas (T. H. Parli-
ment and R. Croteau, eds.), American Chemical Society, Washington, DC, 1986,
p. 34.
2. G. Wasserman, H. Stahl, W. Rehman, and P. Whitman, Kirk-Othmer Encyclopedia
of Chemical Technology, 4th ed., Vol. 6, John Wiley and Sons, New York, 1993,
p. 793.
3. W. Jennings and M. Filsoof, Comparison of sample preparation techniques for gas
chromatographic analysis, J. Agric. Food Chem. 25:440 (1977).
4. C. Weurman, Isolation and concentration of volatiles in food odor research, J. Agric.
Food Chem. 17:370 (1969).
5. H. Sugisawa, Sample Preparation: Isolation and Concentration, in Flavor Re-
search, Recent Advances (R. Teranishi, R. Flath, and H. Sugisawa, eds.), Marcel
Dekker, Inc., New York, 1971, p. 11.
6. S. Risch and G. Reineccius, Isolation of Thermally Generated Aromas, in Ther-
mal Generation of Aromas (T. H. Parliment, R. J. McGorrin, and C.-T. Ho, eds.),
American Chemical Society, Washington, DC, 1989, p. 42.
7. H. Maarse and R. Belz, Isolation, Separation, and Identification of Volatile Com-
pounds in Aroma Research, Akademie-Verlag, Berlin, 1981.
8. C.-T. Ho and C. H. Manley, eds., Flavor Measurement, Marcel Dekker, New York,
1993.
9. P. Schreier, Chromatographic Studies of Biogenesis of Plant Volatiles, Huthig, New
York, 1984.
10. Restek Advantage 4:7 (1993).
11. M. Leahy and G. Reineccius, Comparison of Methods for the Isolation of Volatile
Compounds from Aqueous Model Systems, in Analysis of Volatiles. Methods.
Applications (P. Schreier, ed.), de Gruyter, New York, 1984, p. 19.
12. W. Schultz and J. Randall, Liquid carbon dioxide for selective aroma extraction,
Food Technol. 24:1283 (1970).
13. D. A. Moyler, Carbon dioxide extracted ingredients for fragrances, Perf. Flav. 9:
109 (1984).
14. T. H. Parliment, A new technique for GLC sample preparation using a novel extrac-
tion device, Perf. Flav. 1:1 (1986).
15. T. H. Parliment and H. D. Stahl, Generation of Furfuryl Mercaptan in Cysteine-
Pentose Model Systems in Relation to Roasted Coffee, in Sulfur Compounds in
Foods (C. Mussinan and M. Keelan, eds.), American Chemical Society, Washington,
DC, 1994, p. 160.
16. T. Parliment and H. Stahl, Formation of Furfuryl Mercaptan in Coffee Model Sys-
tems, in Developments in Food Science V37A Food Flavors: Generation, Analysis
and Process Influence (G. Charalambous, ed.), Elsevier, New York, 1995, p. 805.
17. G. B. Nickerson and S. T. Likens, Gas chromatographic evidence for the occurrence
of hop oil components in beer, J. Chromatog. 21:13 (1966).
18. T. Schultz, R. Flath, R. Mon, S. Eggling, and R. Teranishi, Isolation of volatile
components from a model system, J. Agric. Food Chem. 25:446 (1977).
19. C. Aug-Yeung and A. MacLeod, A comparison of the efficiency of the Likens and
Nickerson extractor for aqueous, liquid/aqueous, and lipid samples, J. Agric. Food
Chem. 29:502 (1981).
20. B. Johnson, G. Waller, and A. Burlingame, Volatile components of roasted peanuts:
basic fraction, J. Agric. Food Chem. 19:1020 (1971).
21. S. Chang, F. Vallese, C. Hwang, O. Hsieh, and D. Min, Apparatus for the isolation
of trace volatile constituents from foods, J. Agric. Food Chem. 25:450 (1977).
22. J. de Bruyn and J. Schogt, Isolation of volatile constituents from fats and oils by
vacuum degassing, J. Am. Oil Chem. Soc. 38:40 (1961).
23. J. J. Balboni and W. W. Nawar, Apparatus for direct collection of volatiles from
meat, J. Agric. Food Chem. 18:746 (1970).
24. T. Misharina, R. Golovnya, and I. Beletsky, Comparison of the Efficiency of Isola-
tion of Volatiles from Foodstuffs by Co-Distillation and Likens-Nickerson Meth-
ods, in Developments in Food Sci. V35. Trends in Flavor Research (H. Maarse
and D. van der Heij, eds.), Elsevier, New York, 1994, p. 117.
2
Analysis of Food Volatiles
Using Headspace-Gas
Chromatographic Techniques
Thomas P. Wampler
CDS Analytical, Inc., Oxford, Pennsylvania
I. OVERVIEW
Gas chromatography (GC) involves the analysis of volatile organic compounds,

that is, materials that exist in the vapor phase, at least at the typical GC operating
temperatures between 40 and 300C. Because aroma compounds must, by their
very nature, leave the food matrix and travel through the air to be perceived,
they are generally excellent candidates for analysis by GC. Although many of
these compounds may be solvent extracted, distilled, or otherwise isolated from
the food matrix, it is frequently preferable to take advantage of their volatility
and rely instead on techniques of headspace analysis.
Headspace sampling techniques are frequently divided into three broad cat-
egories: static headspace, dynamic headspace, and purge and trap. In each case,
however, the fundamental principle is the samevolatile analytes from a solid
or liquid material are sampled by investigation of the atmosphere adjacent to the
sample, leaving the actual sample material behind. In static headspace techniques,
a small sample (usually about 1 ml) of the atmosphere around the sample is
injected directly onto the GC column. In dynamic techniques, the organic analytes
from larger samples of the headspace are first concentrated, then transferred to
the GC. Dynamic headspace techniques in their simplest form, then, are just
ways to transfer a headspace sample that is too large to inject directly. The term
dynamic headspace is usually used when referring to the analysis of solid
materials, and the term purge and trap generally refers to the analysis of liquid
samples by bubbling the purge gas through them.
25
26 Wampler
All headspace techniques share certain advantages and considerations.

Chief among these is that the analytes are removed from the sample matrix with-
out the use of an organic solvent, so the resulting chromatogram has no solvent
peak. This may be especially important when the compounds of interest are early
eluters or are, in fact, solvents, and the presence of a solvent peak would both
dilute and mask the analyte peaks. In addition, the effects of sample temperature,
matrix solubility, and the volatility of the analyte are important considerations
in optimizing a headspace assay, whether static or dynamic.
II. STATIC HEADSPACE

A. General Considerations
Even though the actual separation of the analytes in a gas chromatograph does
take place in the vapor phase, most samples are injected as a solution of the
analyte in some volatile solvent. The entire sample, solvent and analytes, vapor-
izes in the hot injection port, and the volatiles formed then proceed to the GC
column. Many compounds, however, exist as gases at the temperature at which
they are being sampled or have sufficiently high vapor pressure to evaporate and
produce a gas phase solution. In these cases, the gas itself may be injected into
the GC instead of a liquid solution, either by syringe or by transferring a known
volume of vapor from a sample loop attached to a valve. The amount of gas that
may be injected into a gas chromatograph is limited by the capacity of the injec-
tion port, the column, and consideration of the increase in pressure and flow in
the injection port caused by a gas phase injection. In practical terms, injections
are almost always in the low milliliter range, with sizes of 0.12.0 ml being
typical. The utility of a headspace injection then depends on whether or not
enough of the interesting analytes exist in a l-ml gas sample to be detected reliably
by GC. Many gas phase analyses are conducted by simple injection, including
quality analysis of hydrocarbon products, natural gas, medical gases, and so on,
and in general analytes present at about one part per million (ppm) may be as-
sayed in a reproducible way using this technique.
The volatile analytes in a gas sample may have always been there, like
argon in air, or they may migrate there from some other source, like air pollutants
from the evaporation of spilled gasoline. The controlled analysis of vapors that
have migrated into an atmosphere from some solid or liquid source forms the
basis of static headspace analysis.
B. Static Headspace Sampling

If a complex material, such as a piece of food, is placed into a sealed vessel and
allowed to stand, some of the more volatile compounds in the sample matrix will
Headspace-Gas Chromatographic Techniques 27
leave the sample and pass into the headspace around it. If the concentration of
such a compound reaches about 1 ppm in the headspace, then it may be assayed
by a simple injection of an aliquot of the atmosphere in the vessel. How much
compound enters the headspace depends on several factors, including the amount
of it in the original sample, the volatility of the compound, the solubility of that
compound in the sample matrix, the temperature of the vessel, and how long the
sample has been inside the vessel. The concentration of the analyte in the heads-
pace also depends, of course, on the volume of the vessel being used. At equilib-
rium, the amount of compound A that has escaped from the sample matrix and
exists in the surrounding atmosphere is just the total amount of A minus the
amount still in the matrix:
AHeadspace ATotal AMatrix
and the partition coefficient is just:
AHeadspace
KA
AMatrix
The amount of A that actually gets into the gas chromatograph depends on what
portion of the total headspace is injected:
Vs
AInjected AH
VT
where Vs is the volume of the syringe injection, VT is the total volume of the
headspace sampling vessel, and AH is the amount of compound A in the total
headspace.
Therefore, the amount of A injected is
Vs
AI (AT AM )
VT
In practice, the food sample is placed into a headspace vial, sealed and
warmed to enhance vaporization of the volatiles, and then allowed to stand for
a period of time to establish equilibrium at that temperature. Once the volatiles
have equilibriated, an aliquot of the headspace gas is withdrawn with a syringe
and injected into the gas chromatograph injection port. As an alternative, the
equilibrated headspace may be allowed to pass through a sample loop of known
volume, which is subsequently flushed into the injection port. Static headspace
analysis has been applied to a wide variety of sample types, including herbs (1)
and fragrances (2).
28 Wampler
C. Advantages of Static Headspace Sampling

Chief among the advantages of static headspace sampling is the ability to analyze
a sample for low molecular weight volatiles without the presence of a solvent
peak. This is especially important because many samples analyzed by static head-
space are actually being assayed for residual solvent content. Packaging, pharma-
ceuticals, and many other processed materials incorporate the use of solvents in
some step of their production, and the amount of those solvents retained in the
finished product must be determined. Since the solvents are determined as ana-
lytes in a gaseous matrix, they are not diluted by a solvent that produces a re-
sponse on the GC detector, so the chromatography is simplified and more sensi-
tive.
In addition to eliminating the solvent peak, static headspace presents a tech-
nique that is easily automated (3,4), making it attractive for sample screening
applications (5). Commercial instruments are available from many suppliers that
automatically warm the sample vials, inject the headspace, and begin the GC
run. These automated systems frequently transfer a measured sample loop full
of headspace to the chromatograph instead of using a syringe. The combination
of careful temperature monitoring, equilibrium time, pressure control of the sam-
ple loop, and automatic injection to the chromatograph provides increased repro-
ducibility over manual attempts at headspace analysis, as well as freeing the
analysts time for other functions.
Additional advantages of the static headspace technique include relatively
low cost per analysis, simple sample preparation, and the elimination of reagents.
Because the sample analytes are not extracted from the sample material using a
solvent, there is no need to deal with solvent reduction by evaporation either into
the air, with its concerns about pollution, or by recondensing.
D. Disadvantages of Static Headspace Sampling

As discussed in Sec. II.B, any static headspace analysis can inject only a fraction
of the compound of interest to the chromatograph, since the concentration in the
headspace is in equilibrium with that still in the sample matrix, and only a portion
of the headspace is withdrawn and transferred. Consequently, for very low levels
of analyte concentration in the original sample material, static headspace tech-
niques may lack the sensitivity required for the determination. Elevating the tem-
perature of the sample generally increases the volatility of the analyte, but most
static headspace instruments have the capability of heating samples only to about
150C.
Analyses at fairly low temperatures also limit the usefulness of static head-
space for analytes with higher boiling points. Many materials that may be ex-
tracted and solvent injected onto a GC column and that may elute well at higher
column temperatures will be poorly represented in a static headspace sample

produced with the sample at a cool temperature. Finally, reproducibility depends
on analyzing a sample after it has reached equilibration, and the time required
to achieve this point may, especially for less volatile compounds, be a drawback
for some analyses.
III. DYNAMIC HEADSPACE

A. General Considerations
As the name implies, dynamic headspace involves moving the analytes away
from the sample matrix in the headspace phase. Instead of allowing the sample
volatiles to come to equilibrium between the sample matrix and the surrounding
headspace, the atmosphere around the sample material is constantly swept away
by a flow of carrier gas, taking the volatile analytes with it. This performs two
functions relative to the concentration of the volatiles. First, it prevents the estab-
lishment of an equilibration state, causing more of the volatile dispersed in the
sample matrix to leave the sample and pass into the headspace. Second, it in-
creases the size of the headspace sample used beyond the limit of the actual
sample vessel. It is not unusual to collect samples using a total volume of 100
ml to 1 liter of headspace, which may result in an essentially quantitative removal
of the volatile analytes from the sample matrix.
To take advantage of this increased amount of volatile analyte, the entire
dynamic headspace sample should be transferred to the gas chromatograph for
a single analysis. This is accomplished by venting the carrier gas of the dynamic
headspace through a collection trap, which retains the organic compounds while
letting the carrier pass through. In this way, the analytes from a large headspace
volume are concentrated in the trap, and a dynamic headspace instrument is fre-
quently called a sample concentrator. Since the sample is being purged with
a flow of carrier and the analytes trapped for analysis, the technique is also fre-
quently called purge and trap. In general, the term purge and trap is used
to refer to liquid samples analyzed by bubbling the carrier through the liquid,
while dynamic headspace is used when the sample material is a solid. In either
case, however, the principle of retaining, or concentrating, the organic analytes
in a trap while venting a large headspace volume is the same. The trapping step
may involve adsorption onto a high-surface area sorbent material or cold trapping
by condensing or freezing the analyte in the trap.
A generalized diagram of a dynamic headspace instrument is shown in Fig.
1. The valve may be a 6-port or an 8-port one, providing for directing flow from
the sample to the trap in one direction and from the trap to the GC in the other.
During sample collection, the headspace gas flows from the sample vessel
through the valve, to the trap and out the vent, while GC carrier flow goes directly
30 Wampler
FIGURE 1 Simplified diagram of a general purge-and-trap/GC system.
through the transfer line to the column. When the valve is rotated, the GC carrier
is diverted through the trap, which is heated rapidly to revolatilize the collected
organics, transferring them to the gas chromatograph for analysis.
Commercial instruments also provide for automatic drying of moisture
from the trap, baking the trap at elevated temperatures between runs, and may
include special options to handle water vapor from liquid samples, cryogenic
ability for cold trapping, cryogenic refocusing on the GC column for sharper
peaks, and automation of multiple samples.
B. Advantages of Dynamic Headspace/Purge and Trap

Dynamic headspace techniques offer many of the same advantages of static head-
space (for a detailed comparison, see Ref. 6), including elimination of the solvent
peak, analysis of just the volatiles, automation, and easy sample preparation. In
addition, the trapping stage of the analysis offers increased sensitivity, permitting
the analysis of volatiles present at the parts per billion (ppb) level routinely.
With careful attention to contaminants and instrument background, it has been
demonstrated that purge-and-trap techniques are capable of routine application
in the parts per trillion (ppt) range (7). Further, sorbents offer some selectivity
within the range of volatiles collected, so it may be possible to select a combina-
tion of sorbent and temperature which permits the collection and concentration
of specific analytes while venting others, thus simplifying the analysis.
C. Disadvantages of Dynamic Headspace/Purge and Trap

Because the instrumentation requires the monitoring of several steps, valving,
heating zones, and so on, purge-and-trap instrumentation is more complex, and
may be more expensive to purchase, than other types of sample introduction.
In addition, again because of the functioning of the instrument, there are many
opportunities for malfunctions, including heater damage, valve leaking, contami-
nation, and cold spots. The sources of error in purge-and-trap instruments have
been reviewed by Washall (8), including sample storage, trap heating effects,
carryover and purging efficiency. Compared to static headspace, purge-and-trap
techniques require a little more time per sample, for purging, trap drying, and
trap transfer, all of which require approximately 15 minutes for a typical analysis.
Some of this time is generally transparent, however, since there is no equilibration
time, and much of the sample processing may be done while the gas chromato-
graph is still analyzing the previous sample.
D. Purging Vessels
1. Liquid Samples
Early applications of purge and trap were targeted at the environmental laboratory
for the analysis of water samples. By purging the water with a flow of helium
and trapping the purged organic pollutants, it was possible to assay analytes such
as solvents at the low ppb and high ppt level routinely. Figure 2 shows a typical
analysis of water for aromatics, present at 20 ppb each. To maximize the surface
area between the water and the bubbles of the purge gas, the gas was forced
through a porous frit at the bottom of the vessel, making a stream of very fine
FIGURE 2 Purge-and-trap analysis of 5-ml water sample containing aromatics at 20 ppb

each.
32 Wampler
bubbles, which then passed through the water, carrying away the volatile organic
contaminants. These fritted vessels work well with clean samples, such as drink-
ing water, but are not ideal for all samples. If the sample contains solid particles,
they may clog the frit, making it difficult to clean, creating carryover and im-
pairing the efficiency. For samples other than clear water, a needle or impinger
arrangement is used (Fig. 3). The purge gas is introduced through a needle or
thin tube, which projects below the surface of the water. While the bubbles are
larger, and therefore the purging is a little less efficient than with a fritted sparger,
the whole system is easier to clean and permits the use of simpler, even disposable
sample vessels. This type of purging vessel is especially well suited for the analy-
sis of foods, which contain many constituents that make samples foam, and al-
most certainly include solids, oils, and other contaminating materials. A further
advantage of the needle or impinger style vessel is that the tube is adjustable, so
that the depth into the liquid sample is variable. For some samples, in fact, it
may be best to have the purge gas enter just above the surface of the liquid,
instead of actually bubbling through it. The purging efficiency is reduced, but if
samples are prone to foaming, this is an intermediate between static headspace
and true purge and trap, which will provide increased sensitivity without contami-
nating the valving of the instrument by having the sample foam over into the
pneumatics.
2. Solid Samples
Solid materials, including soils, polymers, foods, vegetation, and arson debris,
just to name a few, are rarely purged in the kind of vessel used for a liquid
sample, since the frit serves no purpose in these cases and would only be a point
of contamination. Instead, the samples are generally placed into a heated flow-
FIGURE 3 Purging vessels for liquid samples.

through cell (for small samples, just a tube). Flow is brought in at one end, passes
through or around the sample, and exits out the other. For larger samples, includ-
ing whole pieces of fruit, entire containers like cans and bottles, and so on, large
bulk headspace samplers have been developed with internal volumes as large
as 1 liter (Fig. 4). Flow is usually brought in and exits through smaller tubing,
rather than making the large sampler a huge tube, because of sealing consider-
ations.
In general, some additional sampling problems are introduced when using
such large sample containers. These problems should be considered when choos-
ing whether to analyze a portion of a sample in a small tube or the entire sample
in a large vessel. Some analysts are concerned about the representative quality
of a small piece taken from a large sample and therefore choose to analyze the
entire thing, requiring a sample vessel with considerable volume. It must be re-
membered, however, that the larger the fittings used, the more difficult it is to
seal them, and the more likely the sample vessel will leak. Further, if the sampling
FIGURE 4 Sample vessels for dynamic headspace analysis of solid materials.

34 Wampler
is to be done by introducing a flow of carrier gas into the sample chamber, the
entire chamber must be pressurized enough to overcome the back-pressure of
the sorbent trap before any carrier will flow out of the vessel to the trap. In a
small vessel this is not usually a problem, but in a larger vessel, because of the
increased surface area, it becomes much more likely that the seals will leak or
even that the top of the will pop off before flow is established to the trap. In
addition, the time required to sample such a large volume is increased, since
there is mixing and turbulence inside the headspace chamber. Finally, there are
temperature considerations. The larger a sample, the less likely that all parts of
it are at the same temperature and the longer it takes to establish thermal equilib-
rium. Since most sample vessels are heated from the outside, the larger the vessel,
the larger the temperature gradient across it. It may be necessary to monitor the
temperature at both the heater location and at the actual sample location, or at
several sample locations, to have a clear idea of what the actual temperature is
during sampling.
Some of these problems may be avoided by using a vacuum sampling ap-
proach instead of a pressurized sample purge for large vessels. In this way, the
sample vessel does not have to be pressurized to overcome the back-pressure of
the trap, and loss of analytes due to vessel sealing is less likely. Instead, the vent
of the trap is connected to a vacuum pump, and the sample is pulled from the
vessel through the trap, then sent out the pump vent. An inlet tube into the sample
vessel provides for replacement air or sample gas, so that the vessel stays at one
atmosphere throughout the sampling.
E. Trapping
Regardless of the sample vessel typefritted, impinger, thermal desorption tube,
or bulk samplerits function is to remove volatile organics from the sample
matrix and carry them away in the purge gas flow. The carrier gas then proceeds
to the trap, where the volatiles are retained and the carrier is vented. Selection
of the trapping technique and medium depends on several factors, including:
Chemical nature of the analyte
Thermal stability of the analyte
Sorption and desorption characteristics of the sorbent
Breakthrough volume of the analyte on the sorbent
Availability and cost of cryogen
Presence of contaminating materials, including water vapor
1. Sorbent Trapping
Many organic compounds can be removed from a stream of gas by passing them
through a tube packed with a finely divided sorbent material. Because of the high
surface area of the sorbent, the organic vapor is likely to collide with it and may
be adsorbed onto its surface. This is the same principle used for purifying gases
and liquids by forcing them through a filter, frequently filled with activated char-
coal, but in this case the fluid (carrier gas) is discarded and the trapped materials
are the compounds of interest. In an ideal case, the organic volatile is held by
the sorbent at room temperature while other materials pass through, and the ana-
lyte can be desorbed by heating the trap only enough to revolatilize it but not
enough to cause thermal degradation. In fact, this is the case for many organic
compounds, which makes the analysis of water samples for organic pollutants
like solvents very straightforward by purge and trap. Other compounds are not
well sorbed, or behave well only on sorbents that also collect unwanted materials.
Some sorbents are quite stable thermally, whereas others produce artifacts at
desorption temperatures. Some sorbents hold volatiles so efficiently that they
must be heated to quite high temperatures to release them, perhaps causing ther-
mal damage in the process. Part of the method-development stage of any dynamic
headspace technique involves evaluation of the sorbent/analyte interaction and
selection of the best trapping material. It is sometimes necessary to use more
than one sorbent in a trap, especially if a wide range of volatiles is to be trapped,
and some analysts prefer to collect the volatiles by cryogenics onto some inert
surface and eliminate sorbents altogether.
a. Tenax
Most sorbent materials are porous polymers similar to (or identical to) the
kinds of materials used to fill packed GC columns for gas analyses. Tenax
(poly-2, 6-diphenyl-p-phenylene oxide) is perhaps the most widely used, gen-
eral purpose sorbent for dynamic headspace techniques. It is capable of sor-
bing a fairly wide range of organic volatiles, is especially good with aromatics,
may be heated to relatively high temperatures for desorption, and is long
lasting. It is not suitable for very volatile hydrocarbons (pentane and below)
or for small alcohols, which is frequently an advantage. Because it has been used
for dynamic headspace-type analyses for such a long time, there is much
information available in the literature regarding its suitability for particular anal-
yses.
b. Other Sorbents
Although it is sometimes regarded as a universal sorbent, Tenax is not suitable
for every application, and many analysts choose to augment or replace it with
other sorbent materials (9). In an effort to extend the purge-and-trap technique,
the U.S. Environmental Protection Agency (EPA) has devised additional traps,
which use Tenax as the primary sorbent backed by other, more retentive sorbents.
To concentrate on a wide range of volatiles, such as in EPA method 502.2, which
36 Wampler
includes compounds as light as vinyl chloride and as heavy as trichlorobenzene,

the trap specifies Tenax, silica gel, and activated charcoal. As a general rule, the
more retentive the sorbent or the smaller the molecules it is capable of retaining,
the more heat is required to desorb the analytes and regenerate the trap. A particu-
lar problem with activated charcoal, and especially silica gel, is their tendency
to adsorb water, which must be dealt with if it is not to be transferred to the gas
chromatograph.
For the analysis of small molecules by trapping and thermal desorption,
several new sorbent materials have been introduced that provide the retentive
ability of activated charcoal, but collect less water. Graphitized carbon sorbents
(Carbotrap, Carbopack) can collect hydrocarbons larger than propane and release
then thermally. Very small molecules, such as chloromethane, may be trapped
using carbon molecular sieves, which differ from standard, inorganic molecular
sieves in that they are prepared by charring polymers at high temperatures.
These include the various Carbosieves, Carboxen, and Ambersorb materi-
als, with Carboxen-569 in particular reported as having a very low water affin-
ity, increasing the ability to collect small organics without transferring too much
water to the analytical instrument. Various combinations of these sorbents, with
and without Tenax, have been demonstrated to provide both good trapping
efficiency and may be desorbed at relatively high temperatures, producing a
tighter analyte plug transferred to the gas chromatograph. This results in better
chromatographic resolution, particularly for the early eluting peaks in the chro-
matogram.
c. Breakthrough Volume
When a volatile organic compound enters a bed of trapping material in a carrier
gas stream, it may be adsorbed by the packing, but not irreversibly, because it
is important to desorb it later for analysis. Some materials are quite firmly ad-
sorbed and will remain on the surface of the sorbent for a considerable time,
requiring fairly high temperatures (150250C) to remove them. Other com-
pounds are not as well adsorbed, even at room temperature, and will eventually
work their way through the sorbent bed, just as a retained compound works its
way through a GC column. The volume of carrier gas that may be passed through
a trap before a particular analyte leaves the other end of the sorbent bed is called
the breakthrough volume. The breakthrough volume depends on the nature of
the compound, its volatility, the interaction between the compound and the sor-
bent, the amount of sorbent used, and the temperature of the trap. In practice, a
safe sampling volume is used to develop a sampling technique, which is a smaller
volume than the actual breakthrough volume and is reported per gram of sorbent
material.
2. Cryogenic Trapping
a. Advantages
Even well-conditioned solid sorbents exhibit out-gassing at the temperatures re-
quired for thermal desorption of adsorbed compounds. Tenax, for example, pro-
duces aromatic volatiles at temperatures above 180C. For many applications,
the amount of organic material produced from the polymer sorbent may be negli-
gible, but for trace-level applications the presence of background peaks from the
sorbent may be a problem. This is accentuated in the analysis of heavier organics,
since they require a higher desorption temperature to transfer from the trap to the
gas chromatograph. Frequently the desorption parameters become a compromise
between temperatures high enough to desorb the analytes efficiently but low
enough to minimize artifacts. One solution is to eliminate the sorbent altogether
and collect the analytes cryogenically (10).
Liquid nitrogen (boiling point 196) and solid carbon dioxide (boiling
point 79) have both been used to chill traps for cryogenic sample concentra-
tion. Whether one uses liquid nitrogen or carbon dioxide depends on availability,
cost, and the temperature range desired. Although it may seem that CO2 would
suffice for many purposes, the fact is that many analysts find they need tempera-
tures of 100C or colder to collect their analytes efficiently. The pneumatics
involved in delivering liquid nitrogen and CO2 as cryogens are significantly dif-
ferent and generally not interchangeable. Liquid nitrogen is usually used at about
20 psi, while CO2 is supplied at about 900 psi. Further, nitrogen stays as a liquid
when delivered, while CO2 becomes a solid, so the cryogenic wells used as reser-
voirs to cool the trap must be designed differently.
By replacing the trap packing with glass beads, glass wool, or some other
inert material, surface area is provided for the analytes to condense upon during
trapping. When the collection step is complete, the trap need only be heated
enough to volatilize the analyte, since it is not necessary to desorb the compounds
from the surface of a sorbent. This has additional advantages for the collection
of thermally unstable materials, which could decompose at temperatures required
for desorption from a porous polymer or charcoal.
Perhaps the greatest advantage of cryogenic trapping is the ability to tune
the trap to the analytes of interest. By chilling the trap just enough to condense
a particular analyte, other, more volatile compounds may be allowed to pass
through and vent from the system, simplifying the analysis. On the other hand,
since traps may be cooled to temperatures below 180C using liquid nitrogen,
very volatile analytes (with the exception of methane) may be collected, which
would break through ordinary sorbent traps. Some analysts use a combination
of sorbent and cryogenics to extend the range of the sorbent, for example, using
a cryotrap filled with Tenax. For applications needing only sorption, the Tenax
38 Wampler
is used at room temperature. When light hydrocarbons or small alcohols are

needed, the collection temperature is dropped and the sorbent becomes a cold
surface for condensation, just like glass beads in a standard cryotrap.
b. Disadvantages
Although in theory one can tune the trap temperature to collect only the desired
compounds, in practice there may well be compounds that behave similarly to
the analytes of interest and are collected anyway. In general, any compound with
a boiling point higher than that for which the trap collection temperature was
designed will also be trapped. Perhaps the most troublesome is water, since it is
present in many samples and creates significant chromatographic problems. The
point of using cryogenics is to collect at subambient temperatures, so it should
be assumed that if water is present in the sample, it will be condensed or frozen
in the cryotrap.
A second drawback to cryogenic collection is the cost of the additional
instrumentation needed to handle the cryogen, including solenoids capable of
functioning at 180C below zero, control electronics, and the cost of the cryogen
itself. If the trap is filled with glass beads for a clean background, cryogen must
be used for every run. In addition, there is a finite timea few minutes each
runneeded to bring the trap from a rest temperature to the cryogenic tempera-
ture for collection. A final caution involves cold spots. It is important to consider
the effects on the system as a whole of cooling a portion of it to 100C. Even
if the cryogenic trap has its own heater, adjacent portions of the pneumatic path,
especially unheated fittings, will also be cooled, and may warm slowly if not
specifically heated. The longer the trapping time, the more pronounced this effect
becomes, and the more important it is to investigate portions of the flow path
that may be inadvertently cooled, creating a source for subsequent poor chroma-
tography, bleed, inefficient transfer of heavier materials, etc.
F. Water Management
Whether performing purge-and-trap analysis of a water sample, a beverage, or
dynamic headspace of a food material, the sample matrix is likely to contain
substantial amounts of water, much of which may be carried away from the sam-
ple and collected with the analyte compounds. Because the presence of even 1
l of water on a capillary GC column poses a serious analytical problem, it is
important to remove this water one way or another before transferring the trapped
organics to the chromatograph.
There are several approaches to managing water vapor in dynamic head-
space analyses, including selection of a trapping medium that is hydrophobic,
trapping the water independently of the analytes (11), venting the water indepen-
dently, and combinations of these.
One reason for the popularity of Tenax as a sorbent is its low affinity for
water, even if the sample being purged is aqueous, so the purge gas is essentially
saturated. A trapping tube filled with 100150 mg of Tenax will still retain about
1 l of water for each of 40 ml of purge gas used in the process, so a 10-minute
purge cycle at 40 ml/min would deliver about 10 l of water to the trap. Since
Tenax does not adsorb water, however, it is usually enough to pass a source of
dry carrier gas through the trap for a minute or two to vent the water from the
trap without disturbing the organics, which are actually adsorbed onto the surface
of the Tenax. The carbon molecular sieve Carboxen-569 is reported also to be
highly hydrophobic and useful in collection of smaller molecules or in conjunc-
tion with Tenax for a wider-range sorbent trap.
In addition to drying the trap by purging it with a dry carrier, there are
approaches used to prevent the water from reaching the trap in the first place or
to eliminate it from the analytes as the trap is backflushed to the gas chromato-
graph. In the simplified diagram of a purge and trap shown in Figure 1, the
positions marked A, B, and C are possible locations for water-removal devices.
Position A is located just after the sample, before the carrier gas goes through
the valve to the trap, B is located just upstream of the trap, after the valve; and
C is positioned after the trap, just before the gas chromatograph itself. At location
A or B, a device could be added that would remove water vapor from the purge
stream but not affect the organic volatiles, which would still be collected on the
trap. Since A and B are upstream from the trap, the size, volume, and so on
involved here would not affect the quality of the chromatography. A device at
location C, on the other hand, must be designed keeping in mind that the analytes
passing through it are on their way to the GC column and additional volume here
could cause peak broadening.
Two types of devices are in current use to help remove water vapor from
the analytical stream of a purge-and-trap instrument, namely, condensation and
permeation. The condensation units produce an intentional cold spot in the pneu-
matics of the system, providing an area for water to condense out of the carrier.
There is always the concern that less volatile organics will drop out as well, at
least partially, reducing recovery and contaminating subsequent runs. In general,
most of the water present in a purge stream can be removed by passing the carrier
through a zone at about 25C, while many organics, even substituted aromatics
and naphthalene, stay vaporized and pass through to the trap. Some instruments
use a plain piece of stainless steel or nickel tubing at positions A or B to accom-
plish this, whereas some fill the tubing with glass beads to increase the surface
area. These zones should have independent heaters permitting the collected water
to be vaporized and vented from the water trap before the next run, or eventually
the water trap will become saturated and stop functioning. Some instruments
actively cool these zones, using either a cryogen or a Peltier device to increase
the efficiency of the water collection. Care must be taken to control the tempera-
40 Wampler
ture, since the colder the trap, the more likely that compounds other than water
will be condensed in the zone. If the cold spot is placed between the trap and
the gas chromatograph at position C, the effect of either the internal volume or
the temperature on the quality of the chromatography must be taken into account.
For some techniques, a fairly large volume is used, and split capillary chromatog-
raphy is recommended to provide more rapid transfer of the analytes through the
volume, reducing peak broadening. As an alternative, spitless chromatography
may still be performed regardless of the volume of the water trap at C if cryogenic
refocusing is performed at the gas chromatograph connection.
Permeation or diffusion devices eliminate water by having it pass through
the wall of the drying tube while the analyte molecules stay in the carrier stream.
Nafion tubing is quite efficient in its ability to remove water vapor from a gas
stream, partly because its polymer structure includes sulfonic acid groups. A
drying tube made with Nafion is usually a double-walled device, with the analyti-
cal stream passing through the center of the Nafion tube in one direction and a
flow of dry air passing the outside of the Nafion in the other direction. Water is
removed from the inside gas stream by the sulfonic acid groups in the polymer
and transferred through the polymer tube to the dry countercurrent air flow, where
it is removed from the system.
Whichever water-elimination device is used, its effect on the organic ana-
lytes in the carrier stream is increasingly pronounced the more similar the analyte
is to water. Polar compounds, especially small alcohols, are likely to be affected
by diffusion-type dryers, while the higher the boiling point of an organic, the
more likely it is to be slowed in its transfer though a condensation trap.
G. Applications
1. Liquid Samples
Because purge and trap was originally developed for the analysis of volatile or-
ganics in water, it seems a logical extension to apply the technique to food sam-
ples that are largely aqueous. In fact, a significant amount of work has been done
by purge and trap in the analysis of beverages [an excellent early compilation is
by Charalambous (12)], including wine (13,14), beer (15), milk (16), coffee (17),
and fruit juices (1820). There are several important practical considerations
when applying a process designed for the analysis of trace levels of volatiles in
water to samples like fruit juices, including the fact that many of the volatile
constituents are present at levels much higher than those encountered in environ-
mental analyses. Beers and wines have levels of alcohol in the percent range,
rather than the ppb range, so care must be taken to select an appropriate sample
size, and dilution or carrier gas splitting may be required to prevent overloading
the gas chromatograph. In addition, many beverages contain high levels of sugars,
undissolved solids, and oils, which can produce foaming in the sample vessel,
contamination, and carryover to the next run.
For many beverage samples, it is wise to start with a significant dilution
for the first analysis and increase the strength of the sample after an initial evalua-
tion of the chromatography and the behavior of the sample in the instrument.
Fritted purging vessels are frequently a problem, since the bubble size is inten-
tionally small, which accentuates foaming, and the foam can be forced out of
the vessel into the pneumatics of the instrument. Better results are frequently
obtained using an impinger or needle style of sparging arrangement, which results
in a few larger bubbles and reduced foaming. Even here, some samples like fruit
juices can produce large bubbles, which continue up the vessel and into the
plumbing. There are two approaches that can help limit this effect. The incorpora-
tion of a bubble breaker into the top of the purging vessel will prick the bubbles
and allow the fluid to run down the sides of the vessel. The second idea is to
place the sparging gas delivery tube just above the surface of the sample liquid
instead of below it. While this reduces the efficiency of the purging process,
many samples are sufficiently concentrated to provide ample volatiles even if the
liquid is not actively bubbled. A variation of this approach is to take a very thick
liquid such as a juice concentrate and coat a small sample as a film onto the
surface of the purging vessel. This increases the surface area of the sample and
promotes increased recovery from the purge gas without causing foaming. In
such cases, a smaller, concentrated sample may be easier to analyze than a larger,
dilute one.
Figures 5 and 6 show chromatograms of two samples of diet cola. In each
case, 0.5 ml of the soda was diluted to 5 ml, then purged at room temperature
with an impinger to a Tenax trap. Most of the volatiles recovered are citrus oil
constituents, which are used to give colas most of their flavor, with the largest
peak at about 10 minutes being limonine. Considerable variation was found
among various colas, including diet and regular, canned and fountain. The differ-
ences were in both the absolute amount of the oil compounds present and in the
relative amounts of specific peaks, indicating that a variety of citrus oils was
used in flavoring the different beverages.
Vegetable oils (2123), fish oils (24), and other nonaqueous liquids (25)
including emulsions have also been studied using purge-and-trap techniques, es-
pecially for volatile constituents that indicate the freshness, stability, oxidation,
or spoiling of the oils.
2. Semisolid Samples
Some sample materials are neither free-flowing liquids nor stable solids, but
blended to be pastes, gels, spreads, etc. These materials may provide a special
sampling consideration. While liquids may be placed into a small bottle or vessel
42 Wampler
FIGURE 5 Purge-and-trap analysis of 5 ml of a 1/10 dilution of diet cola beverage.
FIGURE 6 Purge-and-trap analysis of a 5-ml sample of a 1/10 dilution of another diet

cola beverage.
to purge and solids (see next section) may be warmed and purged from a tube
or cartridge, semisolids may change their character while being sampled. If the
sample is to be warmed to increase volatility of the analytes, the effect of increas-
ing the temperature on the sample composition must be taken into account. A
material may be fairly solid at room temperature, so thermal desorption in a tube
may seem appropriate, but if it melts and flows at a warmer temperature, the
sample material may run into the pneumatics of the instrument and cause consid-
erable contamination. These samples may be diluted with water and sparged, or
a small sample may be suspended in a large amount of material to provide surface
area for the melted sample to spread onto. Figure 7, for example, shows a 2-mg
sample of toothpaste sampled at 70. Dilution of the toothpaste in water before
running could produce a foaming problem, so the sample was placed in the center
of a glass tube filled with glass wool. There was much more glass wool than
toothpaste, so when the sample was warmed and began to spread, it just migrated
into the glass wool. After a few minutes, the water from the sample was purged
out, and a solid residue of the nonvolatile materials remained in the glass wool
after sampling. Other semisolid foods such as cheeses (26) have been assayed
using dynamic headspace techniques. Figure 8 shows a comparison of Cheddar
and American cheeses analyzed in the same manner as the toothpaste. Thirty-
milligram samples were placed into glass tubes and surrounded by glass wool
to prevent the oils from migrating into the sample concentrator oven.
FIGURE 7 Dynamic headspace analysis of 2-mg sample of toothpaste at 70C.

44 Wampler
FIGURE 8 Comparison of dynamic headspace analyses of 30-mg samples of (top) Ameri-

can cheese and (bottom) Cheddar cheese.
3. Solid Samples
In some respects, solid materials are actually easier to sample by dynamic heads-
pace than are liquids and semisolids. If the sample material is truly solid, and
not likely to melt at sampling temperatures, then a small portion of the material
may be placed into a thermal desorption tube or cartridge, heated, and purged
to the trap. For many foods, warming the sample to 50100C while purging it
enhances recovery of volatiles, but some analytes are temperature unstable and
are better purged at temperatures as low as possible. Dynamic headspace analysis
has been applied successfully to such diverse foods as herbs and spices (27), beet
sugar (28), and canned fish (29). To produce the headspace chromatogram of
raw garlic shown in Fig. 9, a small slice of the garlic was warmed to 70 and
purged for 10 minutes to a Tenax trap. Citrus oils may be purged from the peel
in the same way, as shown in Fig. 10, where a small piece of grapefruit peel was
purged for 10 minutes at 75.
Having a variety of purging devices for sample manipulation makes com-
parisons between different sample types fairly straightforward, as shown in Fig.
11. The top chromatogram shows a purge and trap (impinger vessel) of a diluted
sample of orange soda. The label on the soda indicated that it was created using
only natural flavors. To compare natural orange oil to the chromatogram of the
orange soda volatiles, a small piece of orange peel was thermally desorbed as
for the grapefruit sample shown before. The resulting chromatogram is shown
at the bottom of Fig. 11. Only peaks 1, 2, and 3 are identified (-pinene, -
FIGURE 9 Dynamic headspace analysis of raw garlic at 70C.

46 Wampler
FIGURE 10 Dynamic headspace analysis of grapefruit peel at 75C.
pinene, and limonene, respectively), but the presence of many of the orange oil
peaks in the soda clearly shows that the flavoring is from oranges as the manufac-
turer indicated.
Although small samples generally suffice for dynamic headspace of foods,
it is sometimes preferable to examine a rather large amount of material. It is
possible to make sampling vessels in whatever shape and size is desirable, but
it is important to remember that the larger the sample vessel volume, the more
purging is required to evacuate it. If the sample vessel is to be quite large, it may
be a problem to have it seal properly. Since the purge gas must overcome the
backpressure of the sorbent trap before it can flow through it, the sample vessel
must be pressurized to some extent before the analytes will be transported to the
trap. The larger the vesseland more importantly, the larger the lid that must
be sealedthe more likely it is that the vessel will leak and the sample volatiles
escape to the atmosphere rather than being transferred to the trap. This may be
prevented by using a vacuum pump attached to the vent of the trap to draw the
sample out of the container, rather than trying to pressurize the whole sampling
system. In this way, liter size or larger vessels may be used, with the atmosphere
inside the jar drawn directly to the trap. This approach was used to sample the
atmosphere around whole fruits, producing chromatograms like those shown in
Figs. 12 and 13. The entire banana, kiwi, or other fruit was put into a liter jar
FIGURE 11 Comparison of volatiles purged from (top) orange soda with dynamic heads-
pace of (bottom) orange peel.
48 Wampler
FIGURE 12 Dynamic headspace analysis of one whole banana at room temperature.

Peaks are as follows: 1, butyraldehyde; 2, ethyl acetate; 3, ethyl butyrate; 4, butyl acetate;
5, heptanone; 6, amyl acetate; 7, methyl hexanoate; 8, -pinene; 9, -pinene; 10, limonene;
11, octanol; 12, nonanal; 13, decanal.
FIGURE 13 Dynamic headspace analysis of one whole kiwi; peaks as identified in Fig. 12.
that had a fitting connecting it directly to the inlet of the trap and another that
allowed filtered air into the jar to replace that withdrawn during sampling. The
fruit was allowed to stay in the jar for a period of several days, with a 500-ml
sample of the headspace withdrawn once each day to evaluate the volatiles. A
comparison of the chromatograms obtained over 3 days for a whole lemon is
shown in Fig. 14.
It is sometimes equally important to analyze the materials that come in
contact with foods as well as the food itself. Packaging materials may impart
FIGURE 14 Volatiles collected from one whole lemon after 1, 2, and 3 days.
50 Wampler
odors or flavors to foods by transferring volatile or semivolatile compounds from

the wrapping or container to the food (3032). These compounds may have re-
sulted from the manufacture of the packaging [e.g., solvents, residual monomers
(33) and other low oligomers, plasticizers] or could themselves be contaminants
not intended to be part of the formulation. Sometimes a packaging product ad-
sorbs contaminants from the factory air, oil from processing machines, or other
contaminants and then transfers them to the food, resulting in consumer com-
plaints. Alternatively, the packing may adsorb aroma compounds from the food
instead (34). The packaging materials may easily be assayed using dynamic head-
space techniques, liberating the volatiles from the packaging matrix for analysis.
Figure 15 shows the dynamic headspace analysis of a piece of Styrofoam from
a hot beverage cup. Residual styrene monomer, as well as other organic volatiles,
are clearly present in the cup. These organics may be passed into the food or
beverage that is held the container. Figure 16, for example, shows the results of
a purge-and-trap analysis of a microwaveable soup prepared in the Styrofoam
cup in which it was sold. Here a short, wide-bore column was used, for quick
FIGURE 15 Dynamic headspace of styrene foam cup material.

FIGURE 16 Purge-and-trap analysis of soup made in foam cup in microwave showing

styrene from packaging.
analysis and sensitivity, which shows the presence of styrene monomer, passed
from the cup to the soup. Similar analyses show the same effect for hot coffee
and tea allowed to stand in a Styrofoam cup for 5 minutes before sampling.
H. Quantitative Analysis Using Purge and Trap

From its inception, as a way to determine the levels of organic pollutants in water,
purge and trap has been applied in a quantitative approach. Typical methods
developed by the EPA require rigorous standardization and calculations based
on internal standards. As with similar techniques, the internal standard is added
to the sample matrix just before analysis so that it is processed in the identical
way that the sample volatiles are.
For nonenvironmental samples, the same approach should be used. If a
quantitative determination is to be made on a liquid sample, the internal standard
solution should be miscible with the sample matrix to ensure proper dispersion
and, consequently, identical behavior of the analyte and internal standard vola-
52 Wampler
tiles. For solid materials, it is sometimes difficult to add the internal standard to
the sample matrix without the possibility of the internal standard vaporizing as
the sample is placed into the purging tube or vessel. Addition of the internal
standard solution to the sample tube just before purging reduced the chance that
the internal standard will be preferentially volatilized, as does the selection of
an internal standard of similar volatility to the analyte materials. If the solid
material is a powder, the syringe may be inserted into the center of the sample
plug in the tube and the solution expelled directly into the sample. This approach
has been used successfully for the determination of residual solvents in pharma-
ceuticals (35,36) by dynamic headspace. If the sample is a solid piece, such as
citrus peel or peppercorns, the sample may be placed in the tube and held in
place with a generous quantity of glass wool. The internal standard may then be
injected into the glass wool before thermal desorption. Quantitative procedures
have been developed for a variety of food and packaging analyses, including the
determination of ethylene dibromide in prepared foods (37) and the use of multi-
ple runs to quantitate N-nitrosodimethylamine in baby bottle nipples (38).
Some purge-and-trap instruments have trap injection ports that permit the
injection of the internal standard solution onto the trap directly while the sample
is being purged, which ensures that the whole injection is trapped but does not
compensate for any loss due to purging efficiency, vessel leaking, and so on.
With attention to sampling parameters and the selection of a compatible internal
standard, purge-and-trap analyses can easily provide quantitative results with rel-
ative standard deviations for replicates below 5%.
REFERENCES
1. F. Chialva, G. Gabri, P. A. P. Liddle, and F. Ulian, Qualitative evaluation of aromatic

herbs by direct headspace GC analysis, J. H.R.C. & C.C. 5:182 (1982).
2. E. Jones, M. Davis, R. Gibson, B. Todd, and R. Wallen, Application of headspace
GC to complex liquid samples, Am. Lab. 16:74 (1984).
3. B. Kolb, D. Boege, and L. S. Ettre, Advances in headspace gas chromatography:
instrumentation and applications, Am. Lab. 20:33 (1988).
4. P. Wessels and G. Schwinger, Pressure-balanced headspace sampling using an
electronic-pressure-controlled, on-column injector, LC.GC 13(6):480 (1995).
5. C. A. Weston, R. P. Albert, D. N. Speis, L. Williams, and W. Martin, Screening
environmental samples for volatile organics utilizing a static headspace sampler,
Am. Lab. 24:30 (1992).
6. M. E. McNally and R. L. Grob, A review: current applications of static headspace
and dynamic headspace analysis, Am. Lab. 17:20 (1985).
7. L. Lepine and J. Archambault, Parts-per-trillion determination of trihalomethanes
in water by purge and trap gas chromatography with electron capture detection, Anal.
Chem. 64:810 (1992).
8. J. W. Washall and T. P. Wampler, Sources of error in purge and trap analysis of

volatile organic compounds, Am. Lab. 22:38 (1990).
9. R. D. Barnes, L. M. Law, and A. J. MacLeod, Comparison of some porous polymers
as adsorbents for collection of odor samples and the application of the technique to
an environmental malodor, Analyst 106:412 (1981).
10. H. T. Badings, C. de Jong, and R. P. M. Dooper, Automatic system for rapid analysis
of volatile compounds by purge and cold-trapping/capillary gas chromatography,
H.R.C & C.C. 8:755 (1985).
11. R. Westendorf, Managing water in purge and trap GC/MS, Environ. Lab. 24:36
(1992).
12. Analysis of Foods and Beverages, Headspace Techniques (G. Charalambous, ed.),
Academic Press, New York, 1978.
13. S. Datta and S. Nakai, Computer aided optimization of wine-blending, J. Food Sci.
57:178 (1992).
14. H. Kallio, Method of sensitive analysis of wine headspace volatiles based on selec-
tive capillary column trapping, J. Chromatogr. Sci. 29:438 (1991).
15. A. Kaipainen, A study of the aroma profiles of non-alcoholic beer by thermal desorp-
tion and GC-MS, J. High Resolut. Chromatogr. 15:751 (1992).
16. H. T. Badings, C. De Jong, and R. P. M. Dooper, Rapid analysis of volatile com-
pounds in food products by purge and cold-trapping capillary gas chromatography,
Progress in Flavor Research (J. Adda, ed.), Elsevier Science Publishers B. V., Am-
sterdam, 1984, p. 523.
17. D. Robertys and W. Bertsch, Use of computerized pattern recognition in the analysis
of stress induced changes in coffee aroma, H.R.C. & C.C. 10:244 (1987).
18. G. Blanch, G. Reglera, M. Herriaz and J. Tabera, Comparison of different extraction
methods for the volatile components of grape juice, J. Chromatogr. Sci. 29:11 (1991).
19. R. Marsili, Measuring volatiles and limonene-oxidation products in orange juice by
capillary GC, LC GC 4:358.
20. R. J. Phillips, Routine analysis of volatile components in foods and beverages, Am.
Lab. 24:110 (1992).
21. D. M. Wyatt, Dynamic headspace gas chromatography/mass spectrometry technique
for determining volatiles in ambient stored vegetable oils, J. Chromatogr. Sci. 25:
257 (1987).
22. S. Raghavan, S. Reeder, and A. Khayat, Rapid analysis of vegetable oil flavor quality
by dynamic headspace capillary gas chromatography, J. Am. Oil Chem. Soc. 66:
942 (1989).
23. J. M. Snyder and T. L. Mounts, Analysis of vegetable oil volatiles by multiple head-
space extraction, J. Am. Oil Chem. Soc. 67:800 (1990).
24. T. Hsieh, S. Williams, and W. Vejaphan, Characterization of volatile components
of menhaden fish (Brevortia tyrannus) oil, J. Am. Oil Chem. Soc. 66:114 (1989).
25. J. Merkle and D. Larick, Conditions for extraction and concentration of beef fat
volatiles with supercritical carbon dioxide, J. Food Sci. 59:478 (1994).
26. W. Yang and D. Min, Dynamic headspace analysis of volatile compounds of Ched-
dar and Swiss cheese during ripening, J. Food Sci. 59:1304 (1994).
27. T. Wampler, W. Bowe, and E. Levy, Spitless capillary GC analysis of herbs and
spices using cryofocusing, Am. Lab. 16:25 (1985).
54 Wampler
28. R. Marsili, N. Miller, G. Kilmer, and R. Simmons, Identification and quantitation

of the primary chemicals responsible for the characterization malodor of beet sugar
by purge and trap GC-MS-OD techniques, J. Chromatogr. Sci. 32:165 (1994).
29. B. Girard and S. Nakai, Grade classification of canned pink salmon with static head-
space patterns, J. Food Sci. 59:507 (1994).
30. D. Wyatt, Analytical analysis of tastes and odors imparted to foods by packaging
materials, Plastic Film Sheeting 2:144 (1986).
31. R. Marsili, Testing packaging materials: why and how, Food Product Design (June):
78 (1993).
32. T. Wampler, Automated analysis of volatiles in pharmaceutical environments and
products, Pharm. Manuf. 1:5 (1984).
33. H. Ehret, V. Ducruet, A. Luciani, and A. Feigenbaum, Styrene and ethylbenzene
migration from polystyrene into dairy products by dynamic purge and trap gas chro-
matography, J. Food Sci. 59:990 (1994).
34. J. Konczal, B. Harte, P. Hoojjat, and J. Giacin, Apple juice flavor compound sorption
by sealant films, J. Food Sci. 57:967 (1992).
35. T. Wampler, W. Bowe, and E. Levy, Dynamic headspace analyses of residual vola-
tiles in pharmaceuticals, J. Chromatogr. Sci. 23:64 (1985).
36. J. Washall and T. Wampler, A dedicated purge and trap/GC system for residual
solvent analysis of pharmaceuticals, Am. Lab. 25:20C (1993).
37. S. Moon, T. Healy, A. Lichtman, and R. Porter, An automated method for ethylene
dibromide (EDB) determination in food products, Am. Lab. 18:106 (1986).
38. R. Westendorf, A quantitative method for dynamic headspace analysis using multi-
ple runs, J. Chromatogr. Sci. 23:521 (1985).
3
The Analysis of Food Volatiles
Using Direct Thermal Desorption
Casey C. Grimm, Steven W. Lloyd, and James A. Miller
U.S. Department of Agriculture, New Orleans, Louisiana
Arthur M. Spanier
U.S. Department of Agriculture, Beltsville, Maryland
I. INTRODUCTION
Volatile compounds released from foods are monitored to determine composition,

quality, and safety of the product. The very nature of food, a complex mixture
of proteins, carbohydrates, and fats, results in a continuous change in the forma-
tion of the volatile compounds generated by the food over time. The application
of heat during the cooking process, with variable amounts of oxygen and moisture
present, greatly affects the volatile composition of a food sample. The volatile
composition may become more complex with time as labile compounds react to
form new compounds. The presence of some compounds at concentrations as
low as the parts per billion range can have a major impact on the overall flavor
and acceptability of food. Geosmin and methylisoborneol can be detected in water
at the parts per trillion range (1). For these compounds, the human nose is more
sensitive than current analytical instrumentation.
Volatile analysis of foods is used to determine various properties including
quality, purity, origin, and composition. Due to the relatively low concentrations
of volatile materials that can affect the acceptance or rejection of a food, a proce-
dure is normally employed to extract and concentrate the sample sufficiently prior
to instrumental analysis. The predominant method for analyzing volatiles is gas
chromatography (GC). Typically, compounds need to be delivered to the head
of the GC column in the nanogram range in order to be detected. However, spe-
cialized detectors such as ion trap mass spectrometers (ITMS) or sulfur chemilu-
55
56 Grimm et al.
minescence detectors (SCD) are capable of detecting compounds in the picogram

range.
The analysis of volatiles is generally accomplished by an extraction step,
followed by concentration, chromatographic separation, and subsequent detec-
tion. Well-established methods of analysis include solvent extraction, static and
dynamic headspace sampling, steam distillation with continuous solvent extrac-
tion, and supercritical fluid extraction. An overview of sample preparation meth-
ods is provided by Teranishi (2). The chromatographic profile will vary de-
pending upon the method of sample preparation employed, and it is not
uncommon to produce artifacts during this step (3,4). Thermally labile com-
pounds may decompose in the heated zones of instruments to produce a chro-
matographic profile that is not truly representative of the sample.
A chromatographic snapshot of a food samples volatile composition is
taken at one moment in time in order to compare one sample to the next. The
fewer parameters that are varied, the more likely the analysis will be reproducible.
Hence, the less sample manipulation, the fewer variables in the experiment, the
more likely the results can be repeated. Sample manipulation not only includes
the analytical methodology, but also how the food is cooked and stored.
Direct thermal desorption (DTD) is the technique of sparging the volatiles
from a sample matrix and transferring them directly onto the head of a chromato-
graphic column. The matrix is heated to facilitate the extraction of the volatile
compounds from the sample. A cryofocusing unit, or cold trap, is often employed
to focus the analytes at the head of the column for improved chromatographic
peak shape. This technique allows for the qualitative analysis of volatile com-
pounds with little or no sample preparation. Quantitation of volatiles may be
possible, but is problematic. Variations in purging efficiency, loss of purged vola-
tiles through split/splitless injectors, carryover, and the mechanics of the addition
of an internal standards are some of the problems encountered in quantitative
DTD. Problems with sample carryover have been minimized with the introduc-
tion of newer instrumentation that eliminates transfer lines.
II. HISTORY
In an attempt to facilitate sample preparation methods, early researchers would

unscrew the top of the injection port on a gas chromatograph, remove the liner,
and place a second liner filled with their sample directly into the injection port
(5). The sample would be held in place with a plug of glass wool. The hot injec-
tion port, with its flow of carrier gas, would serve to thermally desorb the volatiles
from the food sample onto a packed column held at room temperature. The vola-
tiles from samples such as peanuts and vegetable oils were analyzed by this
method. In addition to burned fingers, this method had a few drawbacks such
Direct Thermal Desorption 57
as broad peak shapes. Liquid CO2 was used to cool the column to subambient
temperatures, focusing the desorbed volatiles onto the front end of the column,
resulting in enhanced chromatographic separation.
An improvement was made in this approach by moving the sample outside
of the injection port into its own heated block. Grob, Zlatkis, Fisher, and Legendre
developed devices for stripping volatiles from samples and introducing them into
the gas chromatograph (6,7). The external closed-loop inlet device (ECID) was
developed and marketed by Scientific Instrumentation Services (SIS) of River
Ridge, Louisiana (Fig. 1). The apparatus consisted of a heating chamber, a six-
port valve, heated stainless steel tubes, and an electronics unit for controlling the
source block and valve temperatures. A number of these instruments were sold
to researchers primarily in the food industry and are still in use (8).
Figure 1 shows the instrument in the load position. When the valve is
turned, the gas flows are diverted along the dashed lines. Approximately 2 g of
sample can be placed in a glass tube with the sample held in place with glass
wool. Oil samples can be analyzed by putting a few drops of the oil directly onto
the glass wool (9). The tube is then inserted into the sample chamber, which
could be preheated over a wide range of temperatures from ambient to 300C.
FIGURE 1 The external closed-loop injection device is shown in the load position. The
dashed lines represent the flow of gases following sample loading.
58 Grimm et al.
After placing the sample into the sample chamber, the heated valve is rotated to
allow the carrier gas to pass through the sample chamber and to sweep the de-
sorbed volatiles through the valve and into the injection port of the gas chromato-
graph. Since all the purge gas goes onto the column, the carrier flow rate is
equivalent to the purge flow rate. Desorption takes place over a period of 430
minutes. Following the purge, the valve is rotated to the run position; the carrier
gas then bypasses the sample chamber and flows directly into the GC. At this
time a GC temperature ramping program is initiated. The sample tube is then
removed, and a stream of nitrogen is used to sweep out any residual volatiles in
the sample chamber and accompanying transfer lines. This system has been used
to directly desorb the volatiles from a wide variety of food samples including
peanuts, sugar, and meat.
In 1984 the ECID was adapted for use with capillary columns (10). The
capillary columns required lower flow rates and had a lower sample capacity.
The lower flow rates going through the transfer lines resulted in more acute prob-
lems with carryover and moisture. When purging samples with high moisture
content, and using cryofocusing, the capillary columns would freeze up and block
all carrier flow. Carryover from volatiles in the transfer line was observed on
blanks run between samples. The application of the ECID to capillary columns
was not as successful as with the more robust packed columns.
Subsequently, an alternative method of introduction was devised em-
ploying a stainless steel glass-lined tube and with a needle secured to the end
(1113). The device would then, like a syringe, be inserted through a septum
into the GC injection port, analogous to a direct injection, and the carrier gas
diverted by a three-port valve to pass through the sample and purge the volatiles
directly from the sample into the injection port. There they would condense at
the head of a column held at subambient temperature. Originally designed for
the desorption of adsorbent traps, the system was readily amenable for use as a
direct thermal desorption system. These home-built systems have reduced purge/
carrier flow rates, but avoided the problem of contaminated transfer lines with
resultant carryover from one run to the next. As the ECID, the interface with the
split/splitless injectors required that the split flow be capped off to prevent sample
loss. If left open, the purged volatiles would exit the split vent in the same propor-
tions as the carrier gas. Under normal conditions with injections by a syringe,
the sample is volatilized by the injection port and the entire sample is forced
onto the head of the capillary column, with only pure carrier gas exiting the split
vent. However, when the carrier gas is used to first purge the sample, the analytes
are mixed with the carrier gas and any split results in a loss of sample being
deposited on the head of the column. With typical split ratios of 50 :1 to 100 :
1, this limits the quantity of volatiles that can be detected to the parts per thousand
range.
An automated system called the short-path thermal desorber (SPTD) (Fig.

2) was developed jointly by researchers at Rutgers University and Scientific In-
strument Services of Ringoes, New Jersey (an independent company, separate
from the previously mentioned SIS) (14). This instrument is placed on top of the
GC injection port and can be adjusted to allow conventional injections without
disconnecting the unit. The sample is placed in a glass-lined stainless steel tube
and held in place with glass wool. An injection needle is affixed to one end and
the other end is screwed into the SPTD. The injection is made pneumatically,
and the sample is not heated prior to injection. Control of the temperature, desorp-
tion time, and equilibration are provided by an electronic controlling unit. An
automated system has an inherent advantage over manual systems for improved
reproducibility from one run to the next. Quantitative data with a precision of
less than 5% STD has been reported (15).
FIGURE 2 The short-path thermal desorber is here shown in the load position. Following
desorption, valve 1 is closed and valve 2 is opened.
60 Grimm et al.
III. PARAMETERS
The setting of the parameters of direct thermal desorption will directly affect the
desorption efficiency, collection, and quantitative analysis of the sample. Desorp-
tion temperatures must be set high enough to facilitate the stripping of the volatile
compounds, yet not alter the sample or analyte. Desorption times should be suffi-
cient to remove the majority of the volatiles. Purge/carrier gas flow rates should
be sufficient to purge the analyte from the sample but not push the desorbed
volatiles through the cryofocusing zone. GC columns must allow sufficient flow
for efficient desorption of the sample and must be suitable for analysis by direct
thermal desorption.
A. Optimization
Sunesson et al. have performed a multivariant optimization of parameters for the
thermal desorptioncold trapping of volatiles (16). Their conclusions, although
based on the use of a specific instrument, are generally applicable for all direct
thermal desorption devices:
1. Set the cold trap to its lowest setting.
2. The highest temperature possible that does not alter the sample should
be used in the injection block.
3. The lining of the cold trap should be a thick stationary phase.
4. The heating of the temperature ramp rate should be as high as possible.
5. The flow rate should be as high as possible.
B. Sample Constraints
The two major sample limitations to DTD of volatiles are moisture content and
sample size. An excess amount of water will result in a blocked capillary and/
or the extinguishing of a GC detectors flame. When used with packed columns
the moisture contents can be higher. For capillary columns, the upper limit is
approximately 5% moisture in the sample (17). The sample size generally ranges
from 1 to 1000 mg. The lower value results from the mechanical inability to
handle small amounts of material and the amount of analyte being present at a
concentration below the limit of detection. Since the purging efficiency is always
less than 100% and is often less than 10%, a 1-mg sample containing an analyte
present at 100 ppm would result in a maximum of 100 ng of analyte at the de-
tector.
For optimum reproducibility, the more uniform the sample particles, the
more consistent the packing will be. The purge/carrier gas flow through the sam-
ple will contact a similar amount of surface area if the sample particles are uni-
form. The ideal sample for DTD is a thermally inert low-moisture powder with
the analytes present at concentrations between parts per million to parts per thou-
sand. This description sounds like that used to describe typical adsorbent materi-
als like Tenax, Chromosorb, and charcoal. The more the food sample resembles
these adsorbent materials, the fewer problems result from the matrix.
Some samples that have been analyzed by DTD and related techniques are
presented in Table 1. This list contains some liquid samples, which are sparged
TABLE 1 Substances Analyzed

by Direct Thermal Desorption
Sample Ref.
Ajax detergent 24
Aspirin 24
Beef 3
23
Candy 24
Carpets 25
Cheese 26
Coffee 27
24
Coriander fruit 17
14
Glad Wrap 24
Marijuana 14
Onion 28
Peanuts 29
24
Pine needles 24
Plant material 30
Plywood 24
Polypropylene films 17
Rugs 24
Soil 14
Soybean oil 6
Spices 27
Sugar 20
21
Vanilla beans 17
31
Vegetable oil 9
VOCs 18
Wine 19
62 Grimm et al.
and passed through a drying reagent and the analytes deposited directly onto the
head of the column (18,19). Although not truly direct thermal desorption, these
techniques are very similar.
C. Columns
Assuming the sample is amenable to DTD, the internal diameter (I.D.) of the
chromatograph column employed is the most critical factor in determining
the success of a DTD method. A trade-off must be made between sample ca-
pacity, water tolerance, and chromatographic resolution. Packed columns are
superior to capillary columns for sample loading and their ability to handle
moisture, but they provide poor chromatographic resolution. Their high flow
rates are consistent with typical purge rates of 40 ml/min, allowing the injection
ports to deliver all of the purge gas onto the head of the column. When packed
columns or mega-bore glass capillary columns (0.75 mm I.D.) are employed,
even samples with high moisture contents are amenable to direct thermal desorp-
tion. With capillary columns, microgram quantities of water deposited during the
direct thermal desorption may result in blockage of the column during cryofo-
cusing.
A capillary column with a 0.32 mm I.D. has an upper flow rate of approxi-
mately 2 ml/min. Sample loading can be improved and column blockage allevi-
ated by using wide-bore and megabore capillary columns. These columns are the
best for use with direct thermal desorption when analyzing samples with a high
degree of moisture such as meat (3). In going to the larger-diameter capillary
columns with thicker films, a subsequent loss in chromatographic resolution
should be expected. A compromise is reached between the high-capacity packed
columns and the low-load high-efficiency capillary columns with the 0.75-mm-
wide bore and the 0.53-mm megabore open tubular columns. These columns
provide satisfactory chromatography and can handle sample loads in the micro-
gram range. The megabore columns are made of glass, and their installation and
removal can be difficult.
Even with the larger-diameter columns, water may still present a problem.
High amounts of water can compromise the integrity of the columns
stationary phase. As the water enters into the detector, it can quench the
flames on flame ionization (FID) and flame photometric detectors (FPD). The
extinguishing of the detectors can be overcome by using an increased flow
rate of both hydrogen and air. This results in a slight decrease in sensitivity.
This also works well with the sulfur chemiluminescence detectors, which
require higher flow rates and in which hydrocarbons are converted to water
and carbon dioxide, while sulfur-containing compounds produce sulfur mon-
oxide.
D. Split/Splitless Injectors
DTD devices are typically interfaced to the gas chromatograph via split/splitless
injection ports. Split/splitless injectors have been developed to allow injected
volatiles to be concentrated at the head of the capillary column, yet still provide
sufficient gas flow to sweep out the injection port. This is accomplished by alter-
ing the flow through the injector. During the splitless mode (Fig. 3a), the carrier
gas enters at the top of the injector and applies pressure on the volatiles to drive
them into the top of the capillary column. The majority of the carrier gas exits
through the septum purge line at the top of the injector. After sufficient time is
allowed to void the injection volume, the flow is changed so that the majority
of the carrier gas sweeps through the injection liner but exits the sweep vent.
This method works well for normal injections using syringes.
FIGURE 3 Diagram of gas flow of an HP split/splitless injector (a) under normal operation
and (b) under DTD operation.
64 Grimm et al.
Direct thermal desorption devices alter the flow through the injector. The
carrier gas mixed with the analytes now enters the injection port through the
sample needle (Fig. 3b). The majority of the carrier gas and volatiles exit through
the septum purge, with only a fraction of the purged volatiles going onto the
head of the column. For this reason, the split ratio must be decreased as much
as possible, and in some cases the split vent may be capped off. However, this
results in a reduced flow of carrier gas through the injection port and can lead
to carryover between runs and irreproducible results. This is especially true for
volatiles being desorbed from complex food samples.
IV. APPLICATIONS
A comparison of DTD was made with purge and trap (P&T) for analyzing vola-
tiles from samples of beet sugar, roasted peanuts, and grilled ground beef. Ali-
quots from the same sample were used for the comparison. Method parameters
were kept the same with two exceptions: the P&T method used N2 as the purge
gas, while the DTD used helium, and following P&T, the Tenax trap was ther-
mally desorbed at 150C. Samples were purged at temperatures determined ex-
perimentally to be optimal.
A. Experimental
A short-path thermal desorption device (SIS, Ringoes, NJ) was installed on an
HP 5890 Series II gas chromatograph. The capillary column used was a DB-5,
30 m, 0.53 I.D column with a 5 m film (J & W Sci, NJ). Following the injection
port, the capillary was passed through a cryofocusing unit (SGE, Australia). A
thermocouple was attached to the cryofocusing zone and the temperature held at
150C. The gas chromatograph was held at 100C for 10 minutes, then ramped
at 3C/min to 200C. A second ramp of 25C/min was used from 200 to 250C.
The temperature was then held for 5 minutes for a total run time of 50 minutes.
The GC program was initiated at the beginning of the desorption so compounds
not retained would be observed at the detector. The capillary flow rate was 3 ml/
min with the split at 11 ml/min and the septum purge at 3 ml/min. The split/
splitless injector was operated in splitless mode for 4 minutes during the desorp-
tion.
1. Purge and Trap

Sugar crystals were ground to produce free-running crystals. Frozen precooked
beef patties were chopped to a fine meal. Roasted peanuts were chopped to about
1/8-inch particles. Approximately 500 mg of sample was placed in a 50-ml test
tube equipped with a sparge needle. The sample was purged with nitrogen for 4
minutes at a rate of 17 ml/min. An adsorbent trap consisting of 200 mg of Tenax,
held in place with glass wool, was used to collect the volatiles. The sugar was
held at 150C, the beef at 70C, and the peanuts at 120C. These temperatures
have been determined to be the highest possible without significantly altering the
sample. After purging and trapping, the Tenax trap was desorbed at 150C for
4 minutes using the SPTD under the same conditions used for DTD.
2. Direct Thermal Desorption

Aliquots of 500 mg each of the same samples used for purge and trap were taken
and placed in a glass-lined stainless steel tube. Glass wool was used to hold the
sample in place. The sample was purged of atmospheric oxygen for the minimal
setting of 1 second prior to injection. The thermal desorption temperature was
held at ambient for 1 minute following injection. This equilibration period is
needed when switching the carrier gas from the normal operation to pass through
the sample. Upon equilibration, the heated blocks close around the sample tube.
The sample was heated to 150C for sugar, 120C for peanuts, and 70C for
beef. Volatiles were swept directly onto the column and cryofocused at 150C
utilizing liquid nitrogen. After a 4-minute desorption, the liquid nitrogen was cut
off and the cryofocusing zone allowed to rise to the GC oven temperature of
100C.
B. Volatiles from Beet Sugar

Beet sugars are prone to adsorb off-odors as a result of contact of sugar beets
with soil microorganisms that produce potent off-flavors. Odor is a major factor
in quality control of the acceptability of the sugar. The volatile compounds previ-
ously reported in beet sugar are primarily mixtures of short-chain fatty acids,
furanones, aldehydes, and alcohols (20,21). The sample chosen possessed an ex-
ceptionally offensive odor and does not represent a typical chromatographic pro-
file of beet sugars. The volatile composition of the sample is dominated by short-
chain fatty acids and straight-chain aldehydes. Figure 4 shows a comparison of
the beet sugar analyzed by the purge-and-trap method and by direct thermal de-
sorption. The total amount of volatiles loaded onto the column is greater when
using the purge-and-trap method. A likely explanation is that the dense packing
of the sugar in the tube for DTD does not allow the purge/carrier gas to efficiently
desorb the volatiles from the sugar.
Acetic acid is one of the first components to elute (Fig. 4, compound #1)
with a retention time of 6.37 minutes. Propionic, butyric, isovaleric, and hexanoic
acids (compounds 2,3,4, and 5) are observed in both methods. These short-chain
fatty acids are the primary causes of the offensive odor of this sugar. The concen-
66 Grimm et al.
FIGURE 4 GC-FID trace of beet sugar. (Top) Volatiles desorbed by purge and trap: 1,
acetic acid; 2, propionic acid; 3, butyric acid; 4, hexanal; 5, heptanal; 6, methylbutyric
acid; 7, 2-octenal; 8, octanal; 9, nonanal. (Bottom) Volatiles directly desorbed onto the
column.
trations of these compounds are much greater using the P&T method relative to
DTD. In the P&T trace (Fig. 4, top), the straight-chain aldehydes heptanal, octa-
nal, and nonanal are also observed, while only trace levels are observed in the
DTD chromatogram.
C. Volatiles from Peanuts

Figure 5 shows the chromatographic traces of a crushed peanut sample analyzed
by the two different techniques. The upper chromatogram was obtained using
the P&T method, while the bottom chromatogram was run using the DTD
method. Again, the conditions were optimized for DTD and not for P&T.
The volatile composition of roasted peanuts consists of aldehydes, alkylpyr-
azines, furanones, and alcohols (22). Compounds identified by standards and re-
tention times are shown in Figure 5. For the peanuts, the total volatile concentra-
tion is greater for the DTD method than for the P&T method. The relatively
larger chunks of the peanut sample prevent close packing and result in enhanced
desorption efficiency from the sample in the DTD method.
FIGURE 5 GC-FID trace of roasted peanuts. (Top) Volatiles desorbed by purge and trap:
1, pentanal; 2, N-methylpyrrole; 3, hexanal; 4, heptanal; 5, 2,5, and 2,6-dimentylpyrazine;
6, 1-octen-3-ol; 7, methylethylpyrazine; 8; 2-pentylfuran; 9, phenylacetaldehyde; 10, vi-
nylphenol. (Bottom) Volatiles directly desorbed onto the column.
Both methods produced large amounts of acetone, pentane, and acetalde-

hyde, which are unresolved at the front end. The straight-chain hydrocarbons
and aldehydespentane, hexane, heptane, heptenal, octenal, and nonanalare
observed in both chromatograms, with concentrations slightly greater in the DTD
method. The pyrazines are observed in greater concentration in the DTD relative
to the P&T chromatogram.
D. Volatiles from Grilled Beef

Figure 6 shows the chromatographic traces of the volatiles from grilled ground
beef. The volatile profile has been shown to vary with purge temperature (23).
A purge temperature of 70C was selected because protein denaturation has been
shown to occur at higher temperatures (3). The sample was taken from a 4-day-
old refrigerated sample and is typical of samples having undergone meat flavor
deterioration. The large peak at 8.5 minutes is hexanal, which overloads the capil-
lary column in both the P&T trace and the DTD trace. As observed with the
sugar sample, the total volatile concentration is greater in the P&T method.
68 Grimm et al.
FIGURE 6 GC-FID trace of grilled beef. (Top) Volatiles desorbed by purge and trap: 1,
acetaldehyde; 2, pentanal; 3, hexanal; 4, heptanal; 5, 2-octenal; 6, nonanal. (Bottom) Vola-
tiles directly desorbed onto the column.
Pentanal and hexanal (Fig. 6, compounds 2 and 3, respectively) are ob-

served in higher concentrations using the DTD method, while heptanal, 2-octenal,
and nonanal are present in relatively greater concentrations in the P&T method.
These three examples show that for low-boiling compounds, DTD can be
more efficient for desorbing samples, but that the concentrating power of P&T
is needed for higher-boiling compounds. The relative purging efficiency of
P&T versus the desorbing efficiency of DTD is sample dependent.
E. Quantitative Analysis
One of the challenges encountered with direct thermal desorption is in quantita-
tion. With automated instruments, it is fairly easy to reproduce temperatures,
flow rates, and purge times. However, variability can occur as a result of sample
preparation, i.e., granulation, shredding, and chopping, and as a result of sam-
ple packing in the desorption tube. The variation in gas flow through the sample
affects the total amount of material desorbed. The efficiency of the purge/carrier
gas to strip volatiles is directly related to the amount of the samples surface area
with which it comes into contact. The addition of a standard is often employed
to calibrate the purge efficiency and the instrumental response.
There are two types of standards: a surrogate standard and an internal stan-
dard. The surrogate standard is added to the sample prior to any sample manipula-
tion and is used to gauge purge efficiency. These types of standards work well
with liquid matrices where the standard is readily incorporated into the sample
matrix. An internal standard is added to the sample tube prior to DTD and is
used to gauge instrument performance. With little or no sample preparation steps,
the distinction between the two types becomes blurred. The standard itself should
be thermally inert and nonindigenous to the sample. If gas chromatography
mass spectroscopy (GC-MS) is being used, a stable isotope-enriched derivative
of the compound being analyzed is the best standard.
Mechanical problems associated with tube seals and needle blockage may
also cause difficulty with reproducibility. Since the needle remains in the injection
port for several minutes (the injection period plus the desorption period), septa
need to be replaced more frequently than for normal injections. A leak around
the needle will result in a decrease in the sample amount loaded onto the column.
With the SPTD, fast heating of the sample can result in breakthrough at
the cryotrap. A large burst of volatile compounds is blown through the cryotrap
as soon as the heating units are closed. A signal prior to the end of the desorption
period is an indication of the volatiles breaking through the cryotrap. Break-
through is more severe in the case of P&T relative to DTD. A possible explana-
tion is that as the matrix is subjected to heat, the volatiles are more readily de-
sorbed off the Tenax, resulting in a plug of carrier gas containing an increased
concentration of volatiles. A temperature ramp on the heating blocks and/or a
more efficient cold trap could eliminate this problem.
Table 2 shows the averages, standard deviation, and the relative percent
error for a mixture of pentanal, hexanal, heptanal, octanal, nonanal, and decanal
run three times each by DTD. The mixture was desorbed at 150C for 4 minutes
TABLE 2 Repeatability of Three Runs of a 100

ppm Mixture of Hydrocarbons by DTD
Direct Thermal Desorption

Average Std. Dev. %
Pentanal 32923 1831 5.6

Hexanal 41436 2145 5.2
Heptanal 33072 1395 4.2
Octanal 62003 1507 2.4
Nonanal 59425 1513 2.5
70 Grimm et al.
FIGURE 7 GC/MS traces of aromatic rice from three consecutive runs of (top) 875, (mid-
dle) 790, and (bottom) 750-mg samples. 1, Pentane/acetone; 2, acetic acid; 3, pentanal;
4, hexanal; 5, 2-pentalfuran; 6, nonanal.
and cryofocused at 150C. Peak areas were measured using an HP 3390 integ-
rator. The relative percent error for pentanal is 5.6% but decreases to 2.5% for
nonanal.
The chromatographic traces from the DTD-GC-MS of three aliquiots of a
commercial aromatic brown rice are shown in Fig. 7. Three samples consisting
of 0.50 g each of cracked rice were thermally desorbed at 70C. The entire col-
umn was held at 0C during the desorption (4 min) and then ramped at 5C/
min to 200C. No breakthrough was observed, and the first 5 minutes of the
chromatograms have been cut off. The broad peaks at the front end result from the
poor cryofocusing of the low-boiling compounds. The straight-chain aldehydes
resulting from lipid oxidation dominate the chromatogram. Some variability is
observed between runs in the relative concentrations of the aldehydes. The
amounts of acetic acid (2), 2-pentylfuran (5), and nonanal (6) remain constant
between runs. However, the amounts of pentanal (3) and hexanal (4) show a slight
increase between runs. These runs demonstrate the variability of the desorbing
efficiency from a sample as a result of packing efficiency.
V. SUMMARY
Direct thermal desorption provides a rapid technique for the qualitative analysis
of solid samples with little or no sample preparation. Volatiles are thermally
desorbed from the sample and concentrated directly onto the head of a GC col-
umn. Similar chromatographic profiles may be obtained using DTD relative to
P&T. Since the purge and desorption times are concurrent in the DTD method,
analysis times are shorter. The relative purge efficiencies are compound and ma-
trix dependent. DTD in some cases may provide a greater amount of material
for detection. This is especially true for low-boiling compounds with higher vapor
pressures.
Food samples that have moderate moisture content can be analyzed, but
these ultimately require additional steps, which may affect the analysis greatly.
Quantitative analysis is possible but is dependent on the specific analyte and
the matrix. The composition of the sample particles must be uniform, enabling
equivalent packing between runs. Septa need to be examined and replaced more
frequently, and methodology for incorporating an internal standard must be devel-
oped.
ACKNOWLEDGMENTS
The authors would like to thank Mary An Godshall and J. V. Verecellotti for
their editorial assistance in the preparation of this manuscript.
72 Grimm et al.
REFERENCES
1. P. E. Persson, Sensory properties and analysis of two muddy odour compounds,

geosmin and 2-methylisoborneol, in water and fish, Water Res. 14:1113 (1980).
2. R. Teranishi and S. Kint, Sample preparation, Flavor Science, Sensible Principles
and Techniques (T. E. Acree and R. Teranishi eds.), American Chemical Society,
Washington, DC, 1993.
3. A. M. Spanier, C. C. Grimm, and J. A. Miller, Sulfur-containing flavor compounds
in beef: Are they really present or are they artifacts? Sulfur Compounds in Foods
(C. J. Mussinan and Mary E. Keelan, eds.) ACS Symposium Series 564, Washington,
DC, 1994, p. 49.
4. E. J. Block, Flavor artifacts, J. Agric. Food Chem. 41:692 (1993).
5. S. P. Fore, H. P. Dupuy, J. I. Wadsworth, and L. A. Glodblatt, A simplified technique
used to study the shelf life of peanut butter, J. Am. Peanut Res. Educ. Assoc. 5:59
(1973).
6. M. G. Legendre, G. S. Fisher, W. H. Schuller, H. P. Dupuy, and E. T. Rayner, Novel
technique for the analysis of volatiles in aqueous and nonaqueous systems, J. Am.
Oil Chem. Soc. 56:552 (1979).
7. M. G. Legendre, H. P. Dupuy, R. L. Ory, and W. O. McIlrath, Inlet system for
direct gas chromatographic and combined gas chromatographic/mass spectrometric
analysis of food volatiles, U.S. Patent 4,245,494 (1981).
8. J. R. Vercellotti, O. E. Mills, K. L. Bett, and D. L. Sullen, Gas chromatographic
analyses of lipid oxidation volatiles in foods, Lipid Oxidation in Food (A. J. St.
Angelo, ed.), American Chemical Society, Washington, DC, 1992, p. 232.
9. H. P. Dupuy, Analysis of vegetable oils, J. Am. Oil Chem. Soc. 54;10:445449
(1977).
10. H. P. Dupuy, G. J. Flick Jr., M. E. Bailey, A. J. St. Angelo, M. G. Legendre, and
G. Sumrell, Direct sampling capillary gas chromatography of volatiles in vegetable
oils, J. Am. Oil Chem. Soc. 62:1690 (1985).
11. Y. Chen, Z. Li, D. Xue, and L. Qi, Determination of volatile constituents of Chinese
medicinal herbs by direct vaporization capillary gas chromatography/mass spec-
trometry, Anal. Chem. 59(5):744 (1987).
12. P. Werkhoff and W. Bretschneider, Dynamic headspace gas chromatography: con-
centration of volatile components after thermal desorption by intermediate cryofo-
cusing in a cold trap. I. Principle and applications, J. Chromatogr. 483:43 (1987).
13. G. Reglero, M. Herraiz, T. Herraiz and J. Sanz, Capillary gas chromatographic deter-
mination of volatiles in solid matrices by direct introduction using a programmable-
temperature vaporizor, J. Chromatogr. 483:43 (1989).
14. J. J. Manura and T. G. Hartman, Applications of a short-path thermal desorption
GC accessory, Am. Lab. 24(8):46 (1992).
15. P. Werkhoff and W. Bretschneider, Dynamic headspace gas chromatography: con-
centration of volatile components after thermal desorption by intermediate cryofo-
cusing in a cold trap. II. Effect of sampling and desorption parameters on recovery,
J. Chromatogr. 488: (1987).
16. A. L. Sunesson, C. A. Nilsson, B. Andersson, and R. Carlson, Thermal desorption
cold trap-injection in high-resolution gas chromatography: Multivariate optimization

of experimental conditions. J. Chromatogr. 623(1):93 (1992).
17. T. G. Hartman, J. Lech, K. Karmas, J. Salinas, R. T. Rosen, and C. T. Ho, Flavor
characterization using adsorbent trapping-thermal desorption or direct thermal de-
sorption-gas chromatography and gas chromatograph-mass spectrometry, Flavor
Measurement (C. T. Ho and Manley, eds.), Marcel Dekker, Inc., 1993.
18. D. Djozan and Y. Assadi, Optimization of the gas stripping and cryogenic trapping
method for capillary gas chromatographic analysis of traces of volatile halogenated
compounds in drinking water, J. Chromatogr. A 697:525 (1995).
19. C. Garcia-Jares, S. Garcia-Martin, and R. Cele-Torrijos, Analysis of some highly
volatile compounds of wine by means of purge and cold trapping injector capillary
gas chromatography. Application to the differentiation of Rias Baixas Spanish white
wines, J. Agric. Food Chem. 43:764 (1995).
20. R. T. Marsili, N. Miller, G. J. Kilmer, and R. E. Simmons, Identification and quanti-
tation of the primary chemicals responsible for the characteristic malodor of beet
sugar by purge and trap GC-MS-OD techniques, J. Chromatogr. Sci. 32(5):165
(1994).
21. M. A. Godshall, C. C. Grimm, and M. A. Clarke, Sensory properties of white beet
sugars, Proceedings of the 1994 Sugar Processing Conferences, Helsinki, Finland,
1994, p. 312.
22. J. R. Vercellotti, K. L. Krippen, N. V. Lovegreen, and T. H. Sanders, Defining
roasted peanut flavor quality. Part 1. Correlation of GC volatiles with roast color as
an estimate of quality, Food Science and Human Nutrition (G. Charalambous, ed.),
Elsevier Science Publishers, 1992, p. 183.
23. A. M. Spanier, A. J. St. Angelo, C. C. Grimm, and J. A. Miller, The relationship
of temperature to the production of lipid volatiles from beef, Lipids in Food Flavors
(C. T. Ho and T. G. Hartman, eds.), ACS Symposium Series 558, American Chemi-
cal Society, Washington, DC, 1994, p. 78.
24. J. J. Manura, Direct thermal analysis using the short path thermal desorption system,
Short Path Thermal Desorption Note, No. 5 (1991).
25. M. E. Lee, N. E. Takenaka, F. A. Beland, and D. W. Miller, Using direct thermal
desorption to access the potential pool of styrene and 4-phenylcylohexene in latex-
backed carpets, Short Path Thermal Desorption Note, No. 20 (1994).
26. M. C. Vidal-Aragon, E. Sabio, J. Gonzalesz, and M. Mas, Analysis of some highly
volatile compounds of wine by means of purge and cold trapping injector capillary
gas chromatography. Application to the differentiation of Rias Baixas Spanish white
wines. Alimentaria 31(258):25 (1994).
27. J. J. Manura, Direct analysis of spices and coffee, Short Path Thermal Desorption
Note, No. 4 (1990).
28. T. Yuiko and A. Kobayashi, Isolation of volatile components of fresh onion by ther-
mal desorption cold trap capillary gas chromatography, Biosci. Biotech. Biochem.
56(11):1865 (1992).
29. J. V. Vercellotti, K. L. Crippen, N. V. Lovegreen, and T. H. Sanders, Defining
roasted peanut flavor quality. Part 1. Correlation of GC volatiles with roast color as
an estimate of quality, Food Science and Human Nutrition (G. Charalambous, ed.),
Elsevier, 1992, p. 183.
74 Grimm et al.
30. J. L. Esteban, I. Martinez-Castro, and J. Sanz, Evaluation and optimization of the

automatic thermal desorption method in the gas chromatographic determination of
plant volatile compounds, J. Chromatogr. A 657:155 (1993).
31. J. J. Manura, Methodologies for the quantification of purge and trap thermal desorp-
tion and direct thermal desorption analyses, Short Path Thermal Desorption Note,
No. 9 (1991).
32. T. G. Hartman, S. V. Overton, J. J. Manura, C. W. Baker, and J. N. Manoa, Short
path thermal desorption: food science applications, Food Technol. 45(6):104 (1991).
4
Solid-Phase Microextraction for
the Analysis of Aromas and
Flavors
Alan D. Harmon
McCormick & Co., Inc., Hunt Valley, Maryland
I. INTRODUCTION
Every flavor and aroma analysis problem begins with the same question: How
does one select a technique from the myriad well-known isolation methods that
will be best suited to the solution of the current problem? The analytical flavor
chemist is faced daily with the separation and identification of complex mixtures.
These mixtures comprise a wide range of organic chemicals that possess varying
polarities and reactivities, usually occur in trace concentrations, and likely are
included in other complex organic and inorganic matrices. Fortunately, most
aroma chemicals are volatile, and procedures for their isolation from foods and
flavors have been established that take advantage of this volatility. Not so advan-
tageous is the length of time usually required to obtain an isolate that is represen-
tative of the original aroma or flavor of the sample. The selection of steam distilla-
tion, solvent extraction, trapping of the volatiles on adsorbents, or combinations
of these methods with other techniques might require several hours before the
chemist can begin the chromatographic separation. The simple act of isolation
may itself introduce artifacts from impurities in the solvents used, or through
decomposition of the matrix or of the flavor chemicals themselves. An ideal ap-
proach to flavor isolation and analysis would provide an analytical sample whose
composition is identical to the chemical mixture within the matrix, that is free
of solvents and other impurities, and that can be completed within a few minutes
75
76 Harmon
with no intermediate processing of the sample. Solid-phase microextraction ap-

proaches this ideal.
II. WHAT IS SOLID-PHASE MICROEXTRACTION?
Solid-phase microextraction (SPME) is rapidly emerging as a robust technique

for the rapid, solventless extraction or preconcentration of volatile and semivolat-
ile organic compounds in a variety of scientific disciplines. It utilizes a parti-
tioning of organic components between a bulk aqueous or vapor phase and the
thin polymeric films coated onto fused silica fibers in the SPME apparatus. The
technique was first described by Berlardi and Pawliszyn for the analysis of envi-
ronmental chemicals in water (1). In the ensuing years, numerous reports have
appeared in environmental science, forensic science, toxicology, botany, microbi-
ology, entomology, foods and flavors, physical chemistry, and in fundamental
studies of SPME methodology. Since the publication of the first edition of this
volume in 1997, there has been an explosion in the use and reporting of solid-
phase microextraction techniques for the analysis of flavor and aroma com-
pounds. A recent search of the food and flavor related literature for 19992000
provided 62 references from the Food Science and Technology Abstracts, 119
references from the Chemical Abstracts Service, and 224 references from Analyt-
ical Abstracts. Information is readily available on the World Wide Web and ex-
tensive compilations of SPME references are maintained at www.cm.utexas.
edu/brodbelt, http:/ /sciborg.uwaterloo.ca/chemistry/pawliszyn, www.varianinc.
com/csb/gcnotes/spmeindex.html, and www.sigma.com/supelco. New terminol-
ogy [Headspace Sorptive Extraction (HSSE)] has been coined (2), and a related
technique [Stir Bar Sorptive Extraction (SBSE)] has been described (3,4). This
modification of the original coated fiber technology has been commercialized
by Gerstel GmbH [Twister] and it will extend the sensitivity of solid-phase
extractions due to the greater absorbing capacity of the thicker (1 mm) polymer
coating. Despite the recent advances in techniques and technologies, the general
mechanism for use of SPME has not changed.
Solid-phase microextraction techniques are independent of the form of the
matrix; liquids, solids, and gases all can be sampled readily. SPME is an equilib-
rium technique and accurate quantitation requires that the extraction conditions
be carefully controlled. Each chemical component will behave differently de-
pending on its polarity, volatility, organic/water partition coefficient, the volume
of the sample and headspace, the rate of agitation, pH of the solution, and the
temperature of the sample. The incorporation of an internal standard into the
matrix and adherence to specific sampling times will usually result in excellent
quantitative correlations. Since the SPME technique requires no solvents and can
Solid-Phase Microextraction 77
be performed without heating the sample, the formation of chemical artifacts is

greatly reduced, if not completely eliminated.
III. THE SPME DEVICE
Figure 1 describes the apparatus introduced by Supelco (Bellefonte, PA) for man-
ual injections. Similar devices have been designed for automated injection tech-
niques and are available for use with Varian and CTC Analytics autosamplers.
The manual device is essentially a modified syringe having a spring-loaded
FIGURE 1 Graphical representation of a solid-phase microextraction (SPME) device. (Re-

printed with permission from Ref. 10. Copyright 1994 American Chemical Society.)
78 Harmon
plunger and a barrel with a detent to allow the plunger to be held in an extended
position during the extraction phase and during the injection/desorption period.
Also contained within the barrel is a modified 24-gauge stainless steel needle
that encloses another length of stainless steel tubing fitted tightly to a short piece
of solid core fused silica fiber. The bottom portion of the fused silica fiber is
coated with a relatively thin film of any of several pure or mixed stationary
phases. This film serves as the organic solvent for the absorptive extraction
of the volatile compounds from the analytical matrix. The needle functions to
puncture the septa sealing both the sample container and the gc injection port
and to protect the fragile fused silica fiber during storage and use.
SPME fibers are commercially available in several thicknesses and are
coated with polymers ranging from the nonpolar polydimethylsiloxane (PDMS)
to the more polar Carbowax. Combinations of Carboxen, PDMS, Carbowax, and
divinylbenzene copolymers are also available, which provide added benefits for
the extraction of specific compound types and allow a choice of absorption or
adsorption characteristics. For most analyses, especially of volatile aroma com-
pounds, a fiber having a 100-m coating of polydimethylsiloxane is often the
preferred choice. If a more rapid equilibration is needed, a fiber having a 30-m
coating of polydimethylsiloxane might be more appropriate. Fibers with a 7-m
thickness of polydimethylsiloxane chemically bonded to the fused silica support
will work well for samples having high boiling components (e.g., polyaromatic
hydrocarbons) or where higher temperatures might be required to thermally de-
sorb them in the injection port of the gas chromatograph. In general, the fibers
coated with thicker films will require a somewhat longer time to achieve equilib-
rium but will provide higher sensitivity due to the greater mass of the analytes
that can be absorbed.
IV. THE MECHANICS OF THE SPME PROCESS
The process of solid-phase microextraction is illustrated in Fig. 2. A sample is

placed into a vial or other suitable container that is sealed with a septum-type
cap. The fiber should be cleaned before analyzing any sample because the poly-
mer phase can absorb aroma chemicals from the air and produce spurious peaks
in the chromatogram. Cleaning can be done in a few minutes by inserting the
fiber into an auxiliary injection port on a gas chromatograph. For liquid sampling,
the SPME needle pierces the septum and the fiber is extended through the needle
and into the solution. During headspace sampling, the fiber is extended into the
vapor phase above a liquid or solid sample. The SPME apparatus and sample
vial can be supported during the equilibration period by placing them inside a
test tube (18 150 mm or larger) or using another suitable support. Both direct
liquid sampling and headspace techniques often benefit from the addition of salt
FIGURE 2 Sequence of events showing extraction steps and desorption (injection) steps
followed to perform an analysis using SPME. The fiber is inserted directly into a liquid
sample with the subsequent absorption of most of the analyte molecules (small circles)
from the solution. (From Supelco, with permission.)
to the solution, or appropriate adjustment of the pH, which enhance the equilib-
rium of the contained aroma compounds toward the organic phase of the SPME
fiber. Some care must be exercised when penetrating the septa because the needle
point on the SPME device is flat. It might be appropriate to use prepunched septa
both for sealing the sample vials and in the injection port of the gas chromato-
graph. Simply inserting a clean needle from a microliter syringe through the
septum to provide a small hole before inserting the SPME fiber can prevent bend-
ing the needle and destroying the fiber.
A small stirring bar often is used to agitate the solution, which greatly
increases the rate of equilibration (5). After a suitable sampling time (120 min-
utes), the fiber is withdrawn into the needle; the needle is removed from the
septum and is then inserted directly into the injection port of a gas chromatograph
for 12 minutes. The absorbed chemicals are thermally desorbed by the heat of
the injection port and are transferred directly to the column for analysis.
Any manner of injection is suitable for SPME as long as the needle can
be introduced through the septum nut and can be extended into the heated zone
of the injector. Because this technique often involves the preconcentration of
very dilute substances, the split ratio of a split/splitless capillary injection port
should be set to a low value (around 10 :1) so that the benefit of the preconcentra-
tion step is not wasted. A splitless injection mode will transfer more of the ab-
80 Harmon
sorbed material to the analytical column for those applications that require higher
sensitivity. For samples in which the analytes are not at trace levels, split flows
of 20 to 50 mL/min might be a requirement. The use of an injection port liner
with an internal diameter of 1 mm or less is recommended and will provide
sharper peaks and better sensitivity for highly volatile compounds, although com-
pletely satisfactory chromatographic separations and peak shapes can be achieved
using a standard split liner packed with glass wool. Cryogenic cooling of the
column is not necessary for most applications, although some sharpening of early
eluting peaks will result if that capability is available. Care should be taken to
ensure that the upper surface of the glass wool or other packing material used
in the injection liner is below the level of the tip of the SPME fiber when it is
inserted and extended into the injection port. The penalty for extending the fiber
into glass wool is often a broken or damaged fiber.
It was mentioned previously that several types of fibers currently are avail-
able and that they exhibit a certain degree of selectivity. For general usage the
nonpolar thick film fibers will provide good recovery for most aroma compounds.
Polyacrylate fibers are not strictly limited to the absorption of polar molecules,
but they do afford greater sensitivity for the analysis of alcohols, phenols and
TABLE 1 Relative Effectiveness of 100-m Polydimethylsiloxane

and 85-m Polyacrylate Fibers for the Headspace SPME Extraction
of a 10-ppm Mixture of Selected Flavor Chemicals in Water
Relative extraction efficiency calculated
from raw area and normalized to
the n-butanol peak from
polydimethylsiloxane
Flavor compound Polydimethylsiloxane Polyacrylate
Ethyl acetate 4 3.1
n-Butanol 1 3.3
Hexanal 54 48.1
cis-3-Hexenol 6 19.5
Benzaldehyde 22 59
Ethyl caproate 320 170
Limonene 1410 590
Linalool 117 174
Eugenol 32 110
-Ionone 530 505
Dimethyl sulfide 7 6.3
Pyrrolidine 7 7
Pyridine 11 3.7
certain aldehydes when compared to esters and hydrocarbons. Table 1 compares

the relative amounts of a mixture of flavor chemicals extracted from the head-
space above an aqueous solution containing 10 ppm of each compound. For this
example, 0.5 mL of the solution was contained in a 4-mL vial maintained at 55C
without stirring, and without the addition of salt. After an appropriate time to
allow for equilibration of the headspace, the mixture was analyzed by suspending
each fiber above the surface of a solution for 10 minutes before gas chromato-
graphic analysis using a flame ionization detector. The peak areas obtained for
TABLE 2 Relative GC Peak Areas of a Flavor Mixture Obtained by

Direct Split Injection and by Different SPME Sampling Methods
Relative peak area %
Direct SPME liquid SPME headspace

Compound injection sampling sampling
Ethyl acetate 4.4 0.2 1.2
Ethyl butyrate 5.0 2.6 11.5
Limonene 6.4 1.2 2.6
Ethyl caproate 4.3 6.9 8.4
3-Hexenyl acetate 4.3 7.8 12.0
cis-3-Hexenol 4.9 0.3 2.1
Benzaldehyde 5.5 1.1 6.0
Linalool 4.5 1.1 6.0
Diethyl succinate 3.4 0.1 0.1
Neral 2.9 7.0 5.9
2-Methylbutyric acid 2.6 0.1 0.1
-Hexalactone 3.4 0.1 0.3
l-Carvone 4.7 9.6 7.9
Geranial 5.0 13.6 9.7
Anethole 4.8 14.1 5.0
Caproic acid 3.2 0.1 0.1
Phenylethanol 4.9 0.2 0.4
-Ionone 4.3 14.9 8.9
Cinnamic aldehyde 4.6 2.5 0.2
Triacetin 2.1 0.2 0.2
-Decalactone 3.7 8.0 1.5
Heliotropin 2.4 0.5 0.2
Triethyl citrate 2.2 0.1 0.1
Ethyl vanillin 3.3 0.1 0.1
Vanillin 3.0 0.1 0.1
Source: Adapted with permission from Ref. 6. Copyright 1994 American Chemical
Society.
82 Harmon
each compound from both analyses were normalized to the area for the n-butanol
peak extracted with the PDMS fiber. In this way the effectiveness of both the
fibers and the individual components could be measured. For this group of flavor
chemicals, the polyacrylate fiber was more effective for extraction of the alcohols
(butanol, cis-3-hexenol, and linalool), benzaldehyde, and the phenolic compound
eugenol. Also evident are the somewhat lower responses for esters (ethyl acetate
and ethyl caproate), hydrocarbons (limonene), and pyridine. All of the compo-
nents were readily measured using both fibers, however. In a similar study Yang
and Peppard (6) compared the effectiveness of SPME extractions to the direct
injection of a flavor mixture (Table 2). They observed similar selectivities among
the sampling methods and estimated detection limits for the compounds of 0.1
ppb to greater than 1 ppm depending on the extraction efficiencies and detector
responses of the individual flavor components. Their data show essentially no
FIGURE 3 Graphical representation of the relative extraction effectiveness of a series of

carboxylic acids using 100-m polydimethylsiloxane (PDMS) and 85-m polyacrylate
(Acrylate) SPME fibers. The acids were dissolved in water at a concentration of 10 ppm
each and the solutions were sampled using the headspace technique. The effect of dissolv-
ing 25% NaCl in the solutions is also represented. A 3-mL vial containing 1 mL of solution
was used for the experiment and the acids were monitored by capillary GC (30 m 0.25
mm DB-1, 1 film, flame ionization detection, injector temperature 235C, split flow 8
mL/min, fiber desorb time 1.0 min).
response for diethyl succinate, caproic acid, triethyl citrate, vanillin or ethyl vanil-
lin using either method of SPME extraction. A chapter has been included in this
book that describes the various types of fiber material as an aid in the selection
of appropriate fibers for the analysis of specific compound types.
Low-molecular-weight carboxylic acids are difficult to extract from aque-
ous solutions using SPME techniques. Formic through butyric acids are miscible
in water, and even caprylic acid (C8) is soluble to the extent of 68 mg/100 g (7).
The low capacity factors of carboxylic acids to nonpolar phases used in capillary
gc columns lead to severe fronting of acid peaks, which often can be used
to identify their presence in mixtures with other flavor compounds. The same
phenomenon also has an effect on the absorption of acids by SPME phases. It
is possible to enhance their extraction by SPME fibers, however. Figure 3 shows
the relative extraction efficiencies for several carboxylic acids, each at a concen-
tration of 10 ppm in water. The results show the effectiveness of headspace ex-
tractions using both the 85-m polyacrylate and 100-m polydimethylsiloxane
fibers alone, and after the addition of 25% NaCl to the solutions. With the excep-
FIGURE 4 Graph showing the relative extraction differences for carboxylic acids between
headspace and direct immersion SPME techniques. The acids were dissolved in water at
a concentration of 10 ppm each and the solutions were extracted for 5 minutes at 40C,
without stirring, using a 85-m polyacrylate fiber. Both aqueous and 25% NaCl solutions
were examined and the extractions were monitored by capillary GC as in Fig. 3.
84 Harmon
tion of caprylic acid, the polyacrylate fiber is more effective for headspace ex-
tractions. The salt effect is dramatic for carboxylic acids above four carbons in
length.
Differences between headspace sampling and direct immersion sampling
are illustrated in Fig. 4. For this example the polyacrylate fiber was used to extract
aqueous solutions of the carboxylic acids both with and without the addition of
25% NaCl. During the direct immersion sampling, a 2-mL portion of each solu-
tion was maintained at 55C without stirring, and the fiber was inserted into the
solution for 5 minutes. Headspace sampling was continued for 5 minutes at 55C
with the fiber held above the surface of a 1-mL sample contained in a 3-mL vial.
This experiment shows that both headspace and direct immersion sampling are
effective for the extraction of carboxylic acids with more than 3 carbons, but that
the higher the carbon number, the better they can be isolated from the solution.
Irrespective of these apparently successful extractions, however, carboxylic acids
remain difficult to isolate from a mixture of flavor compounds in dilute aqueous
solution.
V. THEORETICAL CONSIDERATIONS
The theoretical aspects of solid-phase microextraction have been well docu-

mented (see, e.g., Refs. 6, 812), and the reader should refer to these references
for a complete discussion. Essentially, the principles affecting extraction of or-
ganic compounds from solutions using SPME are the same factors that control
their partitioning between phases of immiscible liquids in a separatory funnel, a
countercurrent extraction system, a Likens-Nickerson distillation head, or any
other liquid-liquid extraction device. Therefore, the factors affecting efficient ex-
traction by these methods (contact time, efficiency of mixing, pH, salt concentra-
tion, temperature, phase ratios, etc.) also affect the partitioning in SPME extrac-
tions. This is not an unreasonable concept if the SPME fiber is regarded simply
as an immobilized liquid phase in contact with an aqueous solution.
The volume of the fiber coating is small relative to the bulk of the aqueous
phase being extracted, and the mass of analyte absorbed by the coating at equilib-
rium is directly related to its initial concentration in the solution and the distribu-
tion coefficients controlling the equilibria. The principle behind the partitioning
process is the equilibrium established for the analyte(s) between the fiber organic
phase and the solution phase. If you consider the equilibrium expression in effect
for the placement of the fiber in the solution
[X]l [X] f
then
Klf [X ] f /[X ]l
where [X ] is the concentration of the flavor chemical in solution (l) and in the
organic phase of the SPME fiber ( f ). Klf is the distribution coefficient for X
between the liquid phase and the fiber. The amount of material absorbed from
the solution by the fiber can be described by the relationship (6,9)
n C0 VlVf Klf /(Klf Vf Vl)
where n is the amount of compound X absorbed by the fiber, C0 is the initial

concentration of X in the solution, and Vl and Vf are the respective volumes of
the solution and the organic phase of the fiber. In order to obtain a quantitative
extraction (90% of the analyte absorbed by the fiber) the distribution coefficient
of X needs to be about an order of magnitude greater than the volume ratio
Vl /Vf of the system (8). Complete extractions are usually never necessary to obtain
quantitative information, because the amounts of each analyte in the SPME phase
are controlled by the various distribution constants in effect under the experimen-
tal conditions. As long as the conditions are carefully repeated from run to run,
the equilibrium concentrations in the phases also will remain constant.
In the case of headspace SPME extraction, the equilibrium partitioning
occurs between the liquid organic phase of the fiber and the vapor phase above
a liquid or solid sample. The diffusion of analytes to the fiber in the vapor phase
is about four orders of magnitude greater than diffusion in solution (9). The speed
of analysis by headspace sampling should reflect this greater rate. In fact, head-
space SPME can reduce the extraction time from 5 minutes to 1 minute or less
and still maintain high sensitivities for most analytes. Heating the sample greatly
increases the diffusion of analytes into the vapor phase. Zhang and Pawliszyn
have taken advantage of this to effect quantitative recoveries of BTEX com-
pounds from difficult matrices by simultaneously heating the sample and cooling
the fiber (13).
Headspace SPME techniques are capable of extracting components having
boiling points much higher than the sampling temperature. The chromatographic
detection of a series of paraffin hydrocarbons following headspace SPME extrac-
tion at different temperatures is shown in Fig. 5. Considering the mild tempera-
tures utilized, the recovery of n-heptadecane (boiling point 302C) is remarkable.
In fact, a room temperature extraction also provided a significant peak for each
of the nine components. It is not surprising, then, that headspace extractions of
food samples can provide considerable detail about the composition of spices,
herbs and flavors. According to Zhang and Pawliszyn, compounds with Henrys
constants greater than 90 atm cm3 mol1 can be isolated using headspace SPME
86 Harmon
FIGURE 5 Stacked chromatograms showing the GC analysis of a series of n-hydrocarbons

isolated by headspace SPME at various temperatures. The sample was prepared by trans-
ferring 1-l of individual hydrocarbon standards into a 4-mL vial and sealing with a
Teflon-coated septum. After 10 minutes equilibration at the indicated temperature, the
headspace vapor was extracted using a 100-m polydimethylsiloxane fiber and was ana-
lyzed by GC with flame ionization detection. The injector temperature for this sample was
set at 300C, and the fiber was desorbed for 1 minute.
at ambient temperature (9). This would include three-ring polyaromatic hydrocar-

bons with boiling points around 340C, for example.
For compounds that have appreciable water solubility, both the transfer to
the vapor phase and the corresponding decrease in the magnitude of K1f can pre-
vent observing them in chromatograms obtained by SPME extraction. This is not
necessarily a problem during flavor analysis because it becomes possible to ana-
lyze very low concentrations of volatile flavor compounds in the presence of high
concentrations of polar solvents and other less volatile compounds by using
SPME extraction techniques. On the other hand, a complete analysis of every
volatile and semivolatile compound contained within a flavor mixture might not
be possible when using SPME as the only isolation technique.
VI. PRACTICAL APPLICATIONS OF SPME

FOR FLAVOR ANALYSIS
A. Quantitative Analysis
1. The Analysis of Ethanol by Internal Standard Calibration
The SPME process can be readily adapted for the rapid quantitative analysis of
volatile compounds. Ethanol is a common solvent used in flavors and fragrances,
it is miscible with water, has a high vapor pressure, and also has good solubility
in organic solvents. There are several official methods for the quantitative mea-
surement of ethanol in flavors, extracts, and beverages. Some of these include
distillation followed by specific gravity determination of the distillate (15)
(method 28.1.04), direct gas chromatographic analysis after incorporation of a
standard and, perhaps, a carrier solvent such as tetrahydrofuran (15) (method
36.1.01), and dichromate oxidation (15) (method 28.1.07). All of these methods
are accurate, but they can be time consuming, utilize hazardous chemicals, or
perhaps contaminate the gas chromatographic system with nonvolatile sub-
stances. Headspace SPME extraction is another alternative that can be used for
the rapid quantitative determination of ethanol in liquids or aqueous solutions of
solids. Because only volatile materials are isolated from the sample, there can be
no contamination of the chromatographic system with non-volatile compounds.
Figure 6 shows the relative FID responses obtained for an aqueous solution
of ethanol and n-propanol following headspace SPME with 100-m polydimeth-
ylsiloxane and 85-m polyacrylate fibers. A calibration curve for ethanol concen-
trations from 0.1% to 20% by volume is indicated in Fig. 7. Both alcohols have
a greater affinity for the more polar polyacrylate fiber, but the correlation coeffi-
cients for both systems show nearly ideal behavior over the concentration range.
These data were collected manually by transferring 1.0 mL of solution to 4-mL
vials and then extracting the headspace for 5 min at a temperature of 40C. Injec-
tion into the gas chromatograph was completed by desorption for 1 min with an
injection port temperature of 235C and a split ratio of about 10 :1. Samples may
be evaluated simply by introducing an appropriate amount of the internal standard
into the sample, mixing, transferring a suitable amount to another vial, and re-
peating the headspace absorption using the same parameters that were used for
preparation of the calibration curve. The ratio of the area of the ethanol peak to
the area of the n-propanol peak remains consistent from run to run. Linearity is
maintained at concentrations above 20% for ethanol in water, but when analyzing
other flavor mixtures that contain chemicals having a higher affinity for the fiber
phase, ethanol concentrations in this range might saturate the small volume of
the fiber and lead to nonlinear behavior. This would be more likely to occur when
using fibers with thin polymeric filmsfor example, the 7-m polydimethylsi-
88 Harmon
FIGURE 6 Overlaid capillary gas chromatograms showing the response differences of

ethanol and n-propanol toward 100-m polydimethylsiloxane and 85-m polyacrylate
SPME phases. The solution (1.0 mL) was analyzed by headspace SPME in a 4-mL Teflon-
sealed vial. An extraction time of 5 minutes at 40C was used for the experiment.
loxane fiber. When this occurs, it becomes necessary to dilute the solution with
water before the analysis.
It is important to maintain control of extraction times and temperatures
when performing quantitative analyses, but the incorporation of an internal stan-
dard that has characteristics similar to the analyte removes some of the error that
would be associated with an externally calibrated method. More effective control
of the analysis can be assured using automated sampling (16,17). The software
associated with the automated versions of SPME allow precise control of the
absorption and desorption times, a preabsorption delay, and multiple samplings
from each vial using either headspace or direct liquid sampling. They also allow
for the absorption phase of the next extraction to begin before completion of the
chromatographic run of a preceding sample that can greatly increase the
throughput for busy laboratories.
FIGURE 7 A calibration curve for ethanol prepared using n-propanol as an internal stan-
dard at 1.0% and GC analysis following headspace SPME extraction. The upper curve
(solid circles) was obtained with an 85-m polyacrylate fiber and the lower curve (open
squares) was obtained using a 100-m polydimethylsiloxane fiber. The linear correlation
coefficients for ethanol concentrations from 0.1% to 20.0% using both fibers is in excess
of 0.999.
2. Relative Quantitation of Isothiocyanates in a Seafood

Cocktail Sauce
It is not always necessary to determine absolutely the amount of material con-
tained in a sample. It might be sufficient to be able to measure consistently the
relative amount of a component over a period of weeks or months to satisfy
the requirements of a shelf-life study, for example. Horseradish volatiles consist
primarily of allyl and phenylethyl isothiocyanates (14). These compounds pro-
vide the characteristic pungency and flavor of horseradish and impart that charac-
ter to the products into which horseradish is formulated. Seafood cocktail sauces
often use varying amounts of horseradish for these flavor notes. Methods are
often needed to measure volatiles in finished products to determine how much
flavor is being lost through evaporation, chemical conversion, absorption by
packaging materials, or other effects of aging. Accelerated storage conditions can
90 Harmon
provide excellent information if the resources are available to observe and mea-
sure the chemical changes that occur. In that regard, headspace SPME was used
to determine the changes in concentration of the horseradish volatiles in a control
sample of cocktail sauce stored at 4C and in a subsample of the sauce that was
stored at 30C for 5 days. Figure 8 shows the chromatographic differences ob-
served using gc/mass spectrometry to detect and quantify the isothiocyanate com-
ponents. For this analysis, phenyl isothiocyanate was added to the cocktail sauce
samples as an internal standard at a concentration of 10 ppm (1 mg/100 g). The
phenyl isothiocyanate standard was prepared and used as a dilute solution in
triacetin. Each sample was thoroughly mixed with a measured amount of the
standard solution in sealed blender jars, and 5-g portions of the mixture were
transferred to 20-mL headspace vials for SPME extraction GC/MS analysis. Se-
lected ion chromatograms were collected for integration based on the individual
molecular weights of the compounds. These were m/z 99 for allyl isothiocyanate,
m/z 135 for phenyl isothiocyanate, and m/z 163 for phenylethyl isothiocyanate.
The chromatogram in Figure 8a was obtained for the control sample and the
graph in Figure 8b was obtained for the sample subjected to elevated temperature.
Phenyl isothiocyanate (peak 3) produced very stable peak areas for the two analy-
ses, whereas the areas for allyl isothiocyanate (peak 2) and phenylethyl isothiocy-
anate (peak 4) decreased significantly in the heated sample when compared with
the control. A simple calculation showed that allyl isothiocyanate decreased by
approximately 71% during the 5-day test, whereas the less volatile and less reac-
tive phenylethyl isothiocyanate decreased by only 29%. It would be possible to
determine the kinetics associated with these declines by monitoring the relative
concentrations at specific intervals during the shelf-life storage cycle.
3. Measurement of Indole in Dehydrated Yeast

Using Standard Addition Techniques
Solid materials containing aroma compounds present additional challenges to the
analyst. Matrix effects need to be addressed, the aroma components are often
present in encapsulated form, homogeneity issues become important, and equilib-
rium parameters between the solid and gas phases are not as well defined as the
case with liquids. Most of these concerns can be alleviated by converting the
solid material into a solution or suspension before analysis.
Indole is a metabolite of the amino acid tryptophan that occurs naturally
in several yeasts and fungi and contributes to their flavor profile. Above a concen-
tration of about 1 ppm, indole imparts an obnoxious aroma and flavor to dehy-
drated yeast, and to the products in which they are used. Headspace SPME can
be used to determine the amount of indole that is present in powdered yeast using
a standard addition technique. It is sometimes difficult to obtain a sample that
does not contain indole for use as a blank, so the preparation of a standard curve
FIGURE 8 A comparison of the changes observed for the volatile isothiocyanates of
horseradish isolated from (a) a sample of cocktail sauce maintained at 4C and (b) a sub-
sample of the same cocktail sauce held at 30C for 5 days. The volatiles were isolated
from the cocktail sauce samples by headspace SPME extraction using a 100-m polydi-
methylsiloxane fiber after spiking the samples with phenyl isothiocyanate (peak 3, internal
standard) at a concentration of 10 ppm. Analysis of the components was by GC/MS, and
quantitation was performed using selected ion extraction chromatograms of the respective
molecular ions of the isothiocyanates. The other numbered peaks are allyl thiocyanate (1),
allyl isothiocyanate (2), and phenylethyl isothiocyanate (4).
92 Harmon
using an internal standard method may not be a reasonable option. For this type
of analysis it is necessary to control carefully the extraction time, temperature,
and dilution volumes to overcome potentially severe matrix effects.
Carefully measured 2.000-g portions of the powdered yeast are placed into
20-mL headspace vials and diluted with either 10 mL of deionized water or 10
mL of deionized water containing 2 g of indole. The vials are tightly capped
using Teflon-faced septa and then vortexed or otherwise mixed well to suspend
all of the dry material. They should then be placed in a thermostatted oven or
waterbath at about 40C before analysis for indole by headspace SPME GC/MS
using the ion at m/z 117 for quantitation. At least two sets of vials are prepared
for each samplethe first is used to determine the peak area of the indole in the
native material and the second to measure the sum of the peak areas of the
native indole plus the amount of indole added to the sample as a standard. The
difference between the two sets of data can be used to calculate a response factor
for the indole standard, and finally the amount of native indole can be calculated.
Indole in dry yeast can be measured accurately between 50 ppb and 10 ppm using
headspace standard addition methods with a 100-m PDMS fiber.
4. Analysis of Methionyl Acetate in Canary Melon

Using Stable Isotope Dilution Analysis
Stable isotope dilution analysis (SIDA) is arguably the most accurate technique
available for the quantitative determination of organic compounds. It involves the
addition of an isotopically labeled form (either deuterium or carbon-13) of the com-
pound being evaluated to the sample matrix followed by extraction of the mixed
compounds and analysis by GC/MS. The only difference between the standard
and the analyte molecules is an increase in mass corresponding to the number of
deuterium or carbon-13 atoms that have been introduced into its structure. Because
the standard being added to the sample has the same molecular structure as the
compound being analyzed, the technique is not susceptible to matrix effects. Both
the analyte and the standard are exposed to exactly the same environment and be-
have as a single compound during the analysis, including coelution during the chro-
matographic separation. Any force acting on one molecule will act on the other at
the same rate and to the same extent. GC/MS analysis can easily differentiate the
mass differences in the extracted isotopic mixture and allow accurate quantitation
through integration of the respective selected ion chromatograms. The technique
is suitable for use with any compound for which a suitable labeled standard is avail-
able, either commercially or through chemical synthesis.
Stable isotope dilution analysis was used to determine the amount of 3-
methylthiopropyl acetate (methionyl acetate) in the fruit of a ripe canary melon.
A deuterium-labeled standard was prepared from 3-methylthiopropan-1-ol and
acetic anhydride-d6 using standard organic synthesis practices, and it was distilled
to provide a purity of 99.6% measured by capillary gas chromatography. The

mass of this compound (mw 151) is three mass units greater than the natural
compound (mw 148). This is an important consideration because the labeled stan-
dard should have a mass higher than the greatest isotopic peak in the spectrum of
the analyte in order to simplify the calculations after analysis. Because methionyl
acetate contains a single sulfur atom, its mass spectrum contains a substantial
isotope peak at M 2 (m/z 150). The addition of three deuterium atoms in the
SIDA standard extends the mass beyond this isotope cluster.
The purified standard was diluted in acetone to a concentration of 100
g/mL. Ethanol was not selected as the solvent because of the possibility of
degradation of the standard through a transesterification reaction. A 100-g sample
of ripe canary melon fruit was placed into a blender along with 1.0 mL (1 ppm)
of the standard solution. The fruit mixture was pureed and 5-g samples were
removed to Teflon-sealed 20-mL headspace vials for analysis. The volatiles re-
leased from the melon were absorbed onto a 100-m PDMS fiber for 10 minutes
at room temperature and analyzed by GC/MS using selected ion monitoring of
ions 148 and 151. The concentration of methionyl acetate in the melon was deter-
mined as 3.99 ppm after comparing the integration results to a standard curve
prepared in water. Except for the time involved with preparing the isotope-labeled
ester, the complete analysis required little time or effort. The headspace SPME
extraction was efficient, and when coupled with the highly sensitive GC/MS
detector very little sample was required.
One of the benefits of stable isotope dilution analysis is shown in Fig. 9.
The 100-g sample prepared above was kept in a refrigerator and analyzed repeat-
edly over a period of 6 hours. During that time both the methionyl acetate from
the melon and the deuterated methionyl acetate standard decreased in concentra-
tion due to degradation reactions, presumably through enzymatic action. Despite
a reduction of over 70% in the total GC/MS peak area, the calculated concentra-
tion remained constant at a value of 3.99 0.07 ppm.
The quantitation of flavor chemicals using SPME can be rapid, precise,
and accurate if one takes the time to define an appropriate methodology. It is
easy to detect a multitude of volatile chemicals of widely varying classes during a
single analysis, but the incorporation of a single internal standard is not normally
sufficient to quantify every observable component in a mixture. Because SPME
is such a rapid extraction technique, however, it is possible to evaluate the effec-
tiveness of a large number of standards in a relatively short time. One cannot
discount the substantial effects that changes in the matrix might present to your
analysis. One of the most severe is the effect of fats or oils in the system. Fats
compete directly with the small volume of polymeric support available on the
SPME fiber for absorption of the volatile compounds. The effect of the presence
of an increasing amount of milkfat on the accuracy of the techniques described
above is illustrated in Fig. 10. For this example, a solution of ethyl caproate at
94 Harmon
FIGURE 9 Graph showing the decrease in peak area (circles) of methionyl acetate isolated
from a Canary melon puree by headspace SPME versus the amount calculated (triangles)
using stable isotope dilution analysis. A 100-g sample of the puree was spiked with 100
g of deuterated methionyl acetate and 5-g samples were removed from refrigeration at
regular intervals for SPME GC/MS analysis. The mean of the calculated amount was 3.99
0.07 g/g.
a concentration of 10 ppm was calibrated in water, and evaluated in water con-

taining increasing amounts of dairy cream. The accuracy of both internal standard
and external standard calculations begins to fail at a milkfat concentration as low
as 0.1%. Both standard addition and stable isotope dilution analysis measure-
ments did not suffer in this experiment because these techniques are either im-
mune to changes in the matrix (SIDA) or are self-calibrating (Standard Addition).
B. Qualitative Analysis
Solid-phase microextraction is ideally suited to the characterization of unknown
mixtures of volatile organic compounds. As indicated earlier, the different affini-
ties of certain chemical classes to the various polymer films can be used to advan-
tage when applying SPME techniques for their isolation and injection into a gas
chromatograph or GC/MS system. The ability to sample solutions both by immer-
FIGURE 10 Graph comparing the headspace SPME analysis of 10 ppm of ethyl caproate
in solutions of increasing milkfat concentration using four different calibration techniques.
The horizontal dotted line represents the actual concentration of the solutions. The tech-
niques evaluated were standard addition (squares), stable isotope dilution analysis (dia-
monds), internal standard calibration (triangles), and external standard calibration (circles).
The first two points along the X-axis represent solutions containing 0.1% and 0.5% milk-
fat, respectively.
sion and by headspace methods adds another dimension to the isolation technique.
Immersion sampling may provide more sensitivity overall, but headspace extrac-
tion might separate the more volatile materials for increased selectivity. The re-
maining examples will illustrate these concepts more clearly.
1. A Comparison of Direct Split Injection and Headspace

SPME Injection of a Punch Flavor
Liquid flavors are not always simple mixtures of flavor chemicals dissolved in
single, analytically well behaved solvents. Flavorists often combine diluted
forms of chemicals to reach their final goal. Generally ethanol is the solvent of
choice for most applications, but propylene glycol, glycerin, triacetin, benzyl
alcohol, triethyl citrate, fruit juices, sugar syrups, water and other liquids and
solids often find their way into flavor mixtures. A particularly difficult flavor
96 Harmon
FIGURE 11 Chromatogram showing the flame ionization detector response after a direct
split injection of a punch flavor onto a capillary gc column. Analytical conditions: column
(30 meter DB-1, 0.25 mm, 1 film); injector temp. 235C, split flow 100 mL/min; detector
temp (FID) 250C; oven temperature 60 (1 min) to 230 at 4C/min, 0.5 L injected.
problem in this respect is shown in Fig. 11, which shows a capillary gas chromato-
gram of the result of a direct split injection of a fruit punch flavor using flame
ionization detection. Three solvents were used in the flavor, all in large propor-
tion, along with a lesser amount of a fruit juice. Only ethyl caproate could be
identified by gc/ms as a primary flavor chemical among the ethanol, propylene
glycol, and glycerin components. The other, smaller peaks in the chromatogram
were primarily associated with dimeric and polymeric ethers arising from the
solvents. In addition, several artifacts associated with sugar decomposition were
observed.
The same flavor evaluated using the headspace SPME technique is illus-
trated in Fig. 12. Clearly, the two methods provide different results. One should
first compare the contributions of the various solvents to the chromatograms.
Headspace SPME sampling has completely eliminated the glycerin peak, which
revealed 13 additional flavor components that had coeluted with that solvent as
a result of direct split injection. Glycerin could not be detected even using a
selected ion chromatogram. The propylene glycol peak has been reduced to a
well-resolved minor component by the headspace extraction, but ethanol remains
FIGURE 12 Chromatogram showing the FID response to the headspace SPME extract
obtained from six drops of the punch flavor described in Figure 11. The analysis was
performed using a 100-m PDMS fiber exposed to the headspace for 10 minutes at 45C.
The chromatographic conditions were the same as for Figure 11 except that the split flow
was decreased to 8 mL/min and the fiber was desorbed for 1.0 minute in the injection
port.
as a major solvent peak. These differences are due both to the lower volatilities
of propylene glycol and glycerin, and to their hydrophilic nature. The affinity of
the hydroxylic solvents for the polydimethylsiloxane fiber used in this analysis is
much less than the affinities of the less polar, more hydrophobic flavor chemicals.
Although a chromatogram has not been included to show it, a Likens-Nickerson
steam distillation extraction of this flavor also did not provide a satisfactory analy-
sis due to the formation of numerous artifacts from the thermal decomposition
of sugars during the extraction.
2. Isolation of Aroma Volatiles from Fresh Fruits

The natural chemicals comprising the aroma of fresh fruits are usually complex
mixtures of alcohols, aldehydes, esters, and terpenoids that may transform mark-
edly during the ripening cycle. These chemicals are generally recovered from the
fruit pulp or their juices by vacuum or steam distillation before separation and
98 Harmon
analysis using capillary GC or GC/MS. Such isolation techniques require rela-

tively large amounts of fruit, sometimes on the order of several kilograms, to
obtain a suitable analytical sample, which then is diluted and contaminated with
the organic solvents used during the isolation. Additionally, traditional isolation
methods require from 4 to 24 hours before the identification phase can begin.
Purge-and-trap techniques will reduce both the amount of sample required and
the time needed to prepare a suitable isolate, but the equipment is expensive and
requires time to establish the operating interface to a GC or a GC/MS system.
Headspace SPME can be utilized with small samples of fruit, the extract can be
prepared in a few minutes with little sample preparation, and the resulting extract
can be readily transported to any number of gas chromatographic systems.
Yang and Peppard have shown that direct liquid immersion SPME of a
sample of fruit juice beverage was comparable or higher in sensitivity to a con-
ventional solvent extraction using dichloromethane for most of the recovered
flavor chemicals (6). Although they did not compare the sensitivity of headspace
SPME extraction in the same study, it would have provided a similar result. For
studies in our laboratories, fresh fruits have been sampled by simply removing
three or four small cores of fruit pulp using the blunt end of a disposable
Pasteur pipet and depositing the pieces into a 20-mL headspace vial. After fitting
the vial with a Teflon-lined seal, the fruit is immediately extracted using head-
space SPME at room temperature for 10 minutes before injection and analysis
by GC/MS. The result of several of these analyses is shown in Figs. 13 through
15.
Cantaloupe is the orange, delicately flavored fruit of Cucumis melo L.,
which becomes progressively stronger in flavor and aroma with increasing ripe-
ness. Figure 13 shows the result of a GC/MS analysis from a 1.5 g sample of
ripe cantaloupe prepared as described above. The primary aroma compounds have
been identified in the figure. These compounds were not determined in a quantita-
tive experiment, but it has been reported that isobutyl acetate, butyl acetate, and
ethyl butyrate are present in cantaloupe at a level of 0.1 ppm and hexyl acetate
is present at 0.04 ppm (18). Obviously these amounts will vary from sample to
sample, but if these levels are indicative of the concentrations in this melon, the
headspace SPME technique is able to detect very low levels of nonpolar volatile
chemicals. A ripe banana (Musa sapientum L.) was also examined in the same
manner. After transferring about 2-g of banana cores to a vial, the headspace
SPME extract provided the GC/MS chromatogram shown in Fig. 14. This chro-
matogram is considerably more complex than typical banana flavors, which
usually are highly concentrated in isoamyl acetate. Quantitative values have been
reported (18) for several of the compounds from banana extracts, among them
isoamyl alcohol (212 ppm), isobutyl acetate (47 ppm), isoamyl acetate (1275
ppm), isoamyl isobutyrate (0.7 ppm), isoamyl butyrate (6 ppm), isoamyl caproate
(0.07 ppm), eugenol (1.2 ppm), and elemicin (7.5 ppm). The relative peak areas
shown in this chromatogram suggest a different quantitative profile, but this might
FIGURE 13 Total ion chromatogram of a portion of the volatile components obtained

from a sample of ripe cantaloupe by headspace SPME using a 100-m polydimethylsilox-
ane fiber. The fruit was placed into a 20-mL headspace vial by removing small plugs of
the fruit using the blunt end of a disposable pipet. The SPME extraction was carried out
at room temperature for 10 minutes before analysis by GC/MS using the same chromato-
graphic conditions specified for Figure 12 and a Hewlett-Packard 5989 GC/MS system.
The numbered components are ethyl acetate (1), isobutyl acetate (2), ethyl butyrate (3),
butyl acetate (4), 2-methylbutyl acetate (5), hexyl acetate (6), and 2-ethylhexyl acetate
(7).
have been affected by the different ripeness of the samples and the relative extrac-
tion efficiency of the SPME fiber. 3-Hexenyl caproate was found in the SPME
extract but was not listed among the banana flavor compounds in that reference.
A final example from the analysis of fruits is shown in Fig. 15. Figure 15a
represents the volatiles obtained from the flesh of a ripe Bartlett pear (Pyrus
communis L.) prepared in the manner described above. Figure 15b was obtained
by forcing a single pear-flavored jelly bean into a 4-mL vial and acquiring a
headspace SPME sample for 10 min at room temperature. The transfer of the
jelly bean into the vial resulted in the crushing of its outer sugar shell, which
allowed the flavor compounds to escape into the surrounding headspace. With
samples of this type, it is sometimes necessary to introduce water to dissolve the
matrix or the flavor-encapsulating agents before the chemicals are released. It is
evident by comparing the two chromatograms that the aroma chemicals obtained
from the fresh Bartlett pear are different from those observed in the jelly bean
flavor. This is not unusual, because natures biochemists are usually allowed
100 Harmon
FIGURE 14 Total ion chromatogram showing the volatile components obtained from a
ripe banana by headspace SPME using a 100-m polydimethylsiloxane fiber. Small plugs
removed from the banana using the blunt end of a disposable pipet were placed into a
20-mL headspace vial for the extraction. The SPME equilibration time was 10 minutes
at room temperature, and the same chromatographic conditions as listed in Figure 12 were
used. The numbered components are (1) ethanol, (2) isoamyl alcohol, (3) isobutyl acetate,
(4) isoamyl acetate, (5) isobutyl butyrate, (6) isoamyl isobutyrate, (7) isoamyl butyrate,
(8) isoamyl isovalerate, (9) isoamyl caproate, (10) 3-hexenyl caproate, (11) eugenol, and
(12) elemicin.
more freedom when compounding a flavor than are flavor chemists trying to
recreate what nature has provided.
According to the TNO compilation (18) Bartlett pear aroma contains very
low levels of butyl acetate (0.16 ppm), hexyl acetate (0.09 ppm), methyl trans-
2-cis-4-decadienoate (0.05 ppm), ethyl trans-2-cis-4-decadienoate (0.04 ppm)
and -farnesene (0.04 ppm). The concentrations of the other compounds identi-
fied in Fig. 13 were not reported in the reference.
It is apparent from these few examples that SPME can be a powerful tool
for the rapid isolation of volatile chemicals from fresh fruit and fruit juice prod-
ucts. It should be possible to establish a method to determine the degree of ripe-
ness of different fruits by matching the chemical profiles with sensory or other
established parameters. The changes occurring during ripening can be observed
within a few minutes, and the analysis can be repeated using the same piece of
fruit if shallow core samples are taken and proper storage conditions are main-
tained between analyses. A different treatment of the samples using buffers, salt
FIGURE 15 A comparison of the total ion chromatograms of the volatile aroma compo-
nents of (a) a ripe Bartlett pear and (b) a pear-flavored jelly bean isolated by headspace
SPME. Small plugs of the pear were removed with the blunt end of a disposable pipet
and placed into a 20-mL headspace vial for extraction. The jelly bean was forced into a
smaller 4-mL vial. Headspace extraction was performed on each sample for 10 minutes
at room temperature using a 100-m PDMS fiber. Peak identities are as follows: (1) butyl
acetate, (2) hexyl acetate, (3) methyl cis-4-decenoate, (4) ethyl cis-4-decenoate, (5) methyl
trans-2-cis-4-decadienoate, (6) ethyl trans-2-cis-4-decadienoate, (7) -farnesene, (8) isoa-
myl acetate, (9) cis-3-hexenyl acetate, and (10) carveol propionate.
102 Harmon
solutions, homogenization, different SPME fiber types, separation with chiral

capillary columns, and so forth might provide a completely different insight into
the chemical profile and biochemistry of these plants. SPME will not provide a
complete chemical profile for every sample, but as a rapid isolation technique it
can alert the chemist to the types of compounds occurring in the sample during
the time it takes to prepare an extract using a more typical isolation scheme.
3. More Useful Examples of the Headspace SPME of Food

and Beverage Products
For many spices and herbs, aroma and flavor will vary depending on country of
origin, processing conditions, the age of the sample, the type of packaging, the
ratio of essential oil volatiles to heat-producing principles, and many other fac-
tors. Black pepper is one of the most widely consumed spices in the world. It
might be of benefit to be able to evaluate the chemical composition of the volatile
oil of single peppercorns to correlate with sensory attributes. An example of such
an analysis is shown in Fig. 16. This chromatogram was the result of a 5-min
FIGURE 16 Chromatogram showing the volatile compounds isolated from a single black
peppercorn using headspace SPME. Sample preparation included crushing the peppercorn
and quickly placing it into a 4-mL Teflon-sealed vial. The volatiles were collected using
a 100-m polydimethylsiloxane fiber for 5 minutes at room temperature. Chromatographic
conditions were the same as those listed in Figure 12. The peak heights have been increased
by a factor of 5 to show the minor oxygenated terpene components eluting around 20
minutes in the chromatogram.
room temperature headspace SPME extraction of a single peppercorn (42 mg)

that had been crushed and transferred to a 4-mL vial. The chromatogram has
been expanded along the vertical axis to show the smaller components in the
mixture. The composition is somewhat atypical of a normal chromatogram
obtained by steam distillation, but the general appearance is readily identifiable.
Considering that the whole sample could have provided no more than 1.5 mg of
volatile oil, the sensitivity of the extraction is remarkable.
Figure 17 represents the analysis of a sample of curry powder that was
thought to be lacking one of its spice components. Because the spice was known
to contain a unique aroma chemical, it was an easy matter to transfer a small
amount of the curry to a vial, perform a headspace extraction, and determine
whether the spice had been added. The complete analysis required less than one
hour from the time the sample was received in the laboratory. As a quality control
measure, SPME can have a significant impact on the analysis of raw materials
and finished products.
Beverage manufacturers often compete with one another for customer ap-
proval of their flavors. Typical among these were the so-called cola wars of sev-
eral years ago. SPME would have provided a rapid technique to compare various
brands of cola volatiles to determine whether they really were different. The
FIGURE 17 Chromatogram of a portion of the volatile components obtained by headspace

SPME of a sample of curry powder. For this analysis, a 100-mg sample of the dry curry
powder was equilibrated in a 4-mL vial at 55C for 10 minutes and then extracted for 5
minutes using a 100-m PDMS fiber.
104 Harmon
FIGURE 18 Chromatograms showing the similarities between two popular brands of cola
beverages. The volatile compounds were isolated from 1-mL samples of the beverages
using headspace SPME extraction at 55C for 5 minutes with a 100-m polydimethylsilox-
ane fiber.
FIGURE 19 Chromatogram showing the volatile compounds isolated from a sample of

root beer-flavored beverage. The extraction conditions were the same as those described
in Figure 18.
chromatograms shown in Fig. 18 offer a comparison between two cola products.

Obviously these two products are very similar in their aroma profiles, and they
have probably been formulated using similar ingredients. Unfortunately, nothing
can be determined regarding the more subtle volatile compounds or the nonvola-
tile and more polar portions of the beverages using headspace SPME. A much
less complex beverage flavor is shown in Fig. 19. Few people would confuse
the flavor of root beer with a typical cola. Even fewer would find it difficult to
discriminate between their chromatograms.
VII. SUMMARY
During the past few years, solid-phase microextraction has matured as a tool for
the qualitative and quantitative analysis of aroma volatiles. The simplicity of the
technique has made method development strategies a straightforward process for
applications ranging from the complete identification of the volatile components
in a sample to the quantitation of specific chemicals at low part-per-billion levels.
Sustained interest in the process of SPME has resulted in a huge list of publica-
tions in all fields of flavor science during the past several years. As more investi-
gators continue to realize the benefits available from SPME, new applications
will provide insight into areas not yet imagined.
106 Harmon
REFERENCES
1. R. Berlardi and J. Pawliszyn, The application of chemically modified fused silica

fibers in the extraction of organics from water matrix samples and their rapid transfer
to capillary columns, Water Pollution Res. J. Canada 24:179 (1989).
2. C. Bicchi, C. Cordero, C. Iori, P. Rubiolo, and P. Sandra, Headspace sorptive extrac-
tion (HSSE) in the headspace analysis of aromatic and medicinal plants, J. High
Resol. Chromatogr. 23:539 (2000).
3. E. Baltussen, P. Sandra, F. David, and C. Cramers, Stir bar sorptive extraction
(SBSE), a novel extraction technique for aqueous samples: Theory and principles,
J. Microcol. Sep. 11:737 (1999).
4. A. G. J. Tredoux, H. H. Lauer, Th. Heiderman, and P. Sandra, The determination
of benzoic acid in lemon flavored beverages by stir bar sorptive extraction-CGC-
MS, J. High Resol. Chromatogr. 23:644 (2000).
5. C. L. Arthur and J. Pawliszyn, Solid phase microextraction with thermal desorption
using fused silica optical fibers, Anal. Chem. 62:2145 (1990).
6. X. Yang and T. Peppard, Solid-phase microextraction for flavor analysis, J. Agric.
Food Chem. 42:1925 (1994).
7. The Merck Index, 11th ed. (S. Budavari, ed.), Merck & Co., Inc., Rahway, NJ, 1989.
8. D. Louch, S. Motlagh, and J. Pawliszyn, Dynamics of organic compound extraction
from water using liquid-coated fused silica fibers, Anal. Chem. 64:1187 (1992).
9. Z. Zhang and J. Pawliszyn, Headspace solid-phase microextraction, Anal. Chem.
65:1843 (1993).
10. Z. Zhang, M. J. Yang, and J. Pawliszyn, Solid-phase microextraction: a solvent-free
alternative for sample preparation, Anal. Chem. 66:844A (1994).
11. J. Pawliszyn, Solid Phase Microextraction: Theory and Practice, Wiley-VCH, New
York, 1997.
12. Solid Phase Microextraction: A Practical Guide (S. Scheppers Wercinski, ed.), Mar-
cel Dekker, Inc., New York, 1999.
13. Z. Zhang and J. Pawliszyn, Quantitative extraction using an internally cooled solid
phase microextraction device, Anal. Chem. 67:34 (1995).
14. G. Mazza, Volatiles in distillates of fresh, dehydrated and freeze dried horseradish,
Can. Inst. Food Sci. Technol. J. 17:18 (1984).
15. Official Methods of Analysis of AOAC International, 16th ed., Vol. II, Food Compo-
sition; Additives; Natural Contaminants (P. Cunniff, ed.), AOAC International, Ar-
lington, VA, 1995.
16. C. L. Arthur, L. M. Killam, K. D. Buchholz, J. Pawliszyn, and J. R. Berg, Automa-
tion and optimization of solid-phase microextraction, Anal. Chem. 64:1960 (1992).
17. J. R. Berg, Practical use of automated solid phase microextraction, Am. Lab. (Nov.):
18 (1993).
18. Volatile Compounds in Food. Qualitative and Quantitative Data, 7th ed. (L. M. Nijs-
sen, C. A. Visscher, H. Maarse, L. C. Willemsens and M. H. Boelens, eds.), TNO
Biotechnology and Chemistry Institute, Zeist, The Netherlands, 1996.
5
The Advantages of GC-TOFMS for
Flavor and Fragrance Analysis
John F. Holland
Michigan State University, East Lansing, Michigan
Ben D. Gardner
Orion Associates Science & Engineering, Okemos, Michigan
I. INTRODUCTION
As evidenced by the numerous examples in this book, gas chromatography has

evolved into the dominant method for flavor and fragrance analysis. This is to
be expected because the complexity of the samples being analyzed typically man-
dates that some type of separation be achieved before the component analytes
can be measured and characterized, and gas chromatography provides the greatest
resolving power for most of these volatile mixtures. However, even under the
best conditions it is very difficult to resolve all of the components in these com-
plex samples. Because of this, the typical analyst spends considerable time opti-
mizing parameters and developing methods to realize an effective separation of
targeted analytes, where possible. Often overlooked in the method development
process is the relationship between the extent of chromatographic separation re-
quired and the response of the detector. For applications using a single-channel
detector, such as flame ionization detector (FID) or electron capture detector
(ECD), in which the detector response is based on only one shared analyte prop-
erty, adequate qualitative and quantitative characterization of the sample requires
nearly complete and reproducible chromatographic separation of each analyte.
This often requires long elution times or multiple runs under differing column
and instrumental parameters. For analytical veracity, analysis of the peaks present
in the resulting chromatograms requires that they correspond to the retention
107
108 Holland and Gardner
index and area determinants of known standards. Although there are numerous
algorithms to resolve partially coeluting peaks, the analyst must assume that no
minor sample constituents have coeluted with their response hidden under the
peaks of more dominant components. It has been demonstrated that the probabil-
ity of generating coeluting peaks rises dramatically with the number of compo-
nents in a given sample mixture and that the probability of characterizing each
analyte diminishes concomitantly (1). Finally, the presence of unknowns has al-
ways plagued GC analysis, requiring that other methods be used to isolate suffi-
cient amounts of the unknowns for definitive analysis by other analytical tech-
niques.
II. MULTICHANNEL DETECTION
The requirements for complete chromatographic separation can be reduced by

using information obtained from multiple channels using different detectors with
unique selectivities. These multichannel approaches have been helpful in some
cases but, because there are a limited number of individual detectors that can be
used in concert, they have not proved to be a universal solution. Other attempts
to gain multichannel performance involve GC in combination with a detection
device that produces a complete second axis of differentiation. Here the situation
improves dramatically as each channel serves as an independent detector from
which individual chromatograms can be generated, with each one representing
a discrete position on the new axis. As shown in Fig. 1, this provides information
FIGURE 1 Two-dimensional discrimination of analytes. Components A and B are re-

solved only along the second dimension; components C and D are resolved only along
the first dimension.
GC-TOFMS in Flavor Analysis 109
in two dimensions from which the behavior of each analyte can be plotted along
two orthogonal ordinates instead of one ordinate. The area, or field, in which
characterizations can be displayed is now exponentially increased, thereby en-
hancing the probability of isolation. The two ordinate axes, time and detection
channel, are truly orthogonal because the information along each axis is acquired
by totally independent means. Although two components may not be separated
on one axis, they may be completely separated on the second axis. This not only
makes identification and quantitation easier, it reduces the need for complete
temporal separation.
There are several full-axis multichannel detection methods, including ab-
sorbance, fluorescence, and light scatter, where the second axis is wavelength;
magnetic resonance, where that axis is frequency shift; and mass spectrometry
(MS), where that axis is the mass to charge ratio, m/z (2). Of these, the mass
spectrometer is the most powerful multichannel detector for GC, exhibiting the
greatest combination of sensitivity and resolution along the second axis.
In the MS technique, sample molecules enter a vacuum chamber where
they are ionized by one of several methods, the most common being electron
bombardment. The abundance of the ions thus created, which range from molecu-
lar size to small fragment ions, are then determined as a function of their m/z.
The relative abundance of each m/z can be plotted as a histogram, called a mass
spectrum, which is often unique for each compound and serves as a fingerprint
that aids in identification and characterization. These mass spectra can be evalu-
ated to determine the original structure of the analytes and compared with refer-
ence libraries for positive identification, providing an unparalleled qualitative
ability.
In GC-MS, the chromatographic information is reconstructed from the mass
spectral data. A plot of the sum of all m/zs versus time generates a reconstructed
total ion chromatogram (RTIC), as shown in Fig. 2a. This is the chromatogram
most similar to what one would obtain with a single-channel universal detector.
Plots of single m/z responses versus time generate ion chromatograms, as shown
in Fig. 2b. It is the combination of the temporal chromatographic data and the
multichannel mass spectral data that generates the two-dimensional orthogonal
field for the discrimination of temporally unresolved components and aids in their
identification. Not only does this multichannel field assist in component isolation,
the mass spectra from which the reconstructions are made are readily available
to aid in the analysis, as illustrated in Fig. 2c.
Mass spectrometers used for GC detection typically have a mass range of
11000 u. This corresponds to the mass range of virtually all GC amenable ana-
lytes. (1 u corresponds to 1/12 the mass of 12C, which has been assigned the
value 12.000000 by IUPAC convention). The resolving power of the mass spec-
trometer is such that each integer value of m/z can serve as an isolated indepen-
dent channel for chromatographic detection. For convenience of file storage vol-
FIGURE 2 Chromatographic reconstructions of (a) the total ion chromatogram (RTIC),

(b) selected ion chromatograms, and (c) an expanded section of the RTIC with mass
spectra.
ume and interactive interpretation, each m/z along the mass axis is usually
reported in a nominal mass scale, where the nominal mass is the sum of the
protons and neutrons of the atoms in the fragment ion (3). Several vast commer-
cial reference libraries have been developed on this basis (4,5). The exact masses
of the atoms do not lie at these integers, however, but rather exhibit u values
slightly above and below these numbers, so caution should be exercised when
using a nominal mass scale. The difference between the nominal mass and the
exact mass of an ion could result in the m/z of particular ions being misassigned,
depending on the elemental composition and total mass for that ion. This differ-
ence is called the mass defect and becomes more significant at higher m/z. How-
ever, most GC-MS systems can be set to correct for the defects and assign the
appropriate nominal mass number to each m/z.
Although a single-channel detector such as the FID produces an integrated
smooth curve along the time axis, the finite period of time required to collect mass
spectrometric data along the orthogonal (m/z) axis creates an interval between
succeeding spectra that is manifested as a discontinuity in the time axis of the
resulting chromatograms, as shown in the RTIC in Fig. 2c. This discontinuity is
potentially limiting and is a primary consideration in GC-MS methods develop-
ment. It has been recognized that for an isolated GC peak, 1012 data points are
needed to reconstruct the elution profile for precise time measurements and to
gain an accurate area measurement for quantitation (6,7). Where deconvolution
is required, 20 or more data points per eluting peak may be needed (8). These
requirements must be factored against the temporal dimensions of the eluting
components to determine the appropriate spectral generation rate for a given anal-
ysis. For example, where coelutions occur, a peak eluting over a period of 1
minute would require one spectrum every 3 seconds, whereas the same peak
eluting in only 1 second would require an acquisition rate of 20 spectra per sec-
ond. These precision requirements can place a severe onus on the ability of the
mass spectrometer to rapidly produce spectra.
A. Types of Mass Spectrometers

Mass spectrometers are generally classified on the basis of their mass analyzer.
The first commercial GC-MS instruments used magnetic sector analyzers; the
ones in common use today are quadrupole filters, ion traps, and more recently,
time of flight (TOF) instruments. In applications with GC, where the fidelity of
the time axis is essential, the method by which the mass spectral information is
obtained by these analyzers is critical. These can be classified as scanning and
array methods (9). The magnetic sector and quadrupole devices are scanning
instruments that produce mass spectra by sequentially measuring the intensity of
individual m/zs over the range selected. By this method, a full mass spectrum
is obtained over a period of time, which becomes the interval defining the resolu-
tion along the time axis of the resulting chromatograms. The maximum acquisi-
tion rate, while still maintaining adequate sensitivity and resolution, for magnetic
sector instruments is from 1 to 5 spectra/sec, resulting in time intervals of 1.0
to 0.2 seconds. For quadrupoles the maximum rate is from 5 to 10 spectra/sec
with intervals of 0.2 to 0.1 seconds. Scanning over a limited m/z range can im-
prove the resolution on the time axis in cases where mass spectral information
can be sacrificed.
A severe problem occurs with scanning instruments when the scanning

consumes a time that is long in relation to the rate of change in analyte concentra-
tion in the source. In this case, the ion intensities across the m/z range will be
measured at different analyte concentrations, resulting in an incorrect final mass
spectrum. This error is called skewing and can be very detrimental to subsequent
analysis (10). Figure 3 depicts a single-component elution recorded with a scan-
ning instrument at different time intervals along the elution profile and clearly
illustrates the importance of the relationship between the eluting analyte concen-
tration and the moment of individual m/z measurement. The qualitative appear-
ance of the mass spectra are altered across the peak and will be altered in a
different manner in successive runs. This is not a problem when using packed
column chromatography, where a compound may elute from several seconds to
minutes, but it is a problem in capillary chromatography where analyte elution
may take only a few seconds or less.
In contrast, array detectors do not scan but rather measure all of the ions
across the m/z range simultaneously. This method precludes any skewing in the
resulting spectra. The ion trap uses a combination of scanning and array technolo-
gies. It is less susceptible to the effects of skewing than the scanning instruments
because it acquires the complete spectrum in sequential segments with each seg-
ment internally comprising an array, devoid of skewing. The effective maximum
spectral generation rate of the ion trap is from 10 to 15 spectra/sec, producing
FIGURE 3 Adverse effects of elution dynamics on mass spectra collected by scanning

MS.
a resolution along the time axis of 0.1 to 0.067 seconds. The TOF mass spectrom-
eter can also accommodate full range array detection because all the ions present
in the source are simultaneously extracted and subsequently measured, producing
spectra completely without skewing (9). Additionally, because no scanning action
is involved, the time required to produce a mass spectrum is greatly reduced. The
maximum rate with adequate sensitivity is from 50 to 200 spectra/sec, providing
resolutions along the time axis of 0.02 to 0.005 seconds. In cases where time
resolution priorities exceed those of the limit of detection, TOF acquisition rates
can be increased to 500 complete spectra per second with a consequent resolution
of 0.002 seconds on the time axis (1113). The ability to resolve along the time
axis is essential for GC-MS and the rate at which spectra can be generated at
acceptable sensitivity and mass resolution is critical. Clearly, as the eluting peaks
from the chromatograph become sharper, the necessity for speed on the mass
axis becomes more demanding and the abilities of array detection become more
attractive.
III. TECHNIQUES OF GC-MS
The parameters for the operation of the mass spectrometer in GC-MS are primar-
ily determined by the time restraints imposed by the chromatography and the
limit of detection (LOD) required by the analyst. In the early days of this technol-
ogy, packed columns were almost exclusively employed and the constraints of
time were minimal. This enabled full m/z range scanning to be employed, the
speed of which was dictated by the LOD requiredthe slower the scan, the
more sensitive the measurements. In this mode, the mass spectrometer operates
continuously through the complete elution duration, generating a three-dimen-
sional mass spectral file, the temporal dimensions of which establish the chro-
matographic time axis for subsequent reconstructions. Qualitative confirmations
and the determinants for quantitation are derived from the mass spectra, while
the elution behavior of the various components is derived from the ion chromato-
grams that also provide the area measurements required for quantitative analysis.
The advent and proliferation of capillary columns imposes severe con-
straints on the abilities of the mass spectrometer. Where required, both the resolu-
tion on the time axis and the sensitivity of ion measurement can be improved in
scanning instruments by measuring over only a part of the m/z range. This can
be useful where the nature of the components is known and the range collected
can be set to accommodate their presence (14). In situations where the compo-
nents are not known, the range being collected can be set to vary, moving higher
on the m/z scale as the elution time increases (15). Since larger molecular weight
compounds typically elute later than small ones and often produce larger frag-
ment ions, the amount of spectral information that is lost in these techniques can
sometimes be minimized.
Where lower limits of detection are required and the mass spectra and elu-
tion behavior of the components are known in advance, scanning instruments use
a technique known as selected ion monitoring (SIM) (14). In this approach, only
a limited number of preselected m/zs are measured and stored during the elution
process. This allows relatively more time to be spent measuring each m/z of
interest, increasing the sensitivity with which they are measured. The limited
number of m/zs being monitored reduces the time for each measurement
cycle, increasing the resolution along the chromatographic axis. However, in this
technique, enhancements in chromatographic resolution and sensitivity are ob-
tained at the cost of spectral information. Overall, this is a popular technique for
targeted compound analysis using scanning-based mass spectrometers. In con-
trast, array mass spectrometers, because they simultaneously detect all the ions
within the selected mass range, achieve SIM equivalent detection limits all the
time.
The proliferation of applications and the increasing number of samples ex-
perienced in many areas of GC-MS analysis can rapidly saturate existing re-
sources, creating a demand for increased throughput. As the elution (separation)
time is often a major part of the overall analysis, the quest for more rapid chroma-
tography has been quite active. In one approach, short, narrow-bore, thin phase
capillary columns, along with rapid heating and high carrier gas flow rates, are
used with standard instrumentation to minimize elution times in a technique
called time-compressed chromatography (TCC) (16). This technique significantly
reduces the elution time for complex mixtures, creating a situation where the
occurrence of coelutions increases. As a consequence, high spectral generation
rates and sophisticated deconvolution routines are required to analytically resolve
each component. Typically, minimum acquisition rates of 3050 spectra/second
are required, precluding the use of scanning mass spectrometers in this technique.
A related technique called fast chromatography (17) pushes the time resolution
envelope even further by employing custom sample containment, cyro-concentra-
tion, and ballistic heating to produce narrow peaks 100300 msec wide that re-
quire spectral generation rates of 200 Hz or more. Only TOFMS appears to qual-
ify as an adequate MS detector for these chromatographic advances.
Another GC-MS technique of interest in the analysis of complex samples
is two-dimensional (2-D) gas chromatography, in which the sample is passed
through two columns of different polarity. Although not a complete orthogonal
technique, the differing behavior in these columns provides another dimension
for discrimination, yielding plots of the eluting times from the different columns
on two time axes. Typically, 2-D chromatography is accomplished by a method
called heart-cutting whereby timed selected portions of the primary elution are
diverted to the second column to resolve components that would otherwise co-
elute from the first column. The qualitative power of the mass spectrometer com-
plements the objectives of this approach by providing structural information for
the various components in the mixtures and, because resolution on the time axis
for this approach is not critical, scanning mass spectrometers can readily be em-
ployed.
A newer method of 2-D chromatography is called comprehensive 2-D (18).
In this approach, the two differing columns are connected in series and the entire
sample passes through both columns. A thermal modulator periodically cools and
heats the junction of the two columns. The components in the eluent from the first
column are absorbed at the junction during the cooling phase and then desorbed,
refocused, and passed on to a second, much shorter, narrow-bore column by rapid
heating. This process generates information for construction of three-dimensional
plots of the entire 2-D separations as shown in Fig. 4. The mass spectrometer
adds a fourth dimension and again is extremely useful in component identification
of these complex mixtures (19,20). Mass spectra are acquired in a cycle timed
with the heating process, which is adjusted to allow complete analyte transmis-
sion through the short second column. This requires fine resolution along the
time axis because the heating/cooling cycle is usually in the range of 4 to 10
seconds and the rapid absorption/desorption creates narrow elution peaks of only
a few hundred milliseconds. Although the number of components eluting in each
cycle is limited, in order to gain full range, mass spectra defining this region, in
excess of 100 spectra/second, may be required, again well within the range of
TOFMS but precluding effective use of other instruments.
As an added advantage, this technique of rapid alternate thermal focusing
and release of the analytes flowing in the carrier stream fragments the component
elution from the first column into a series of sharp peaks exiting from the second
column. This behavior increases the signal to noise ratio of the subsequent mea-
surements, and in many cases summing these individual peaks can significantly
improve the limit of detection of the analysis.
FIGURE 4 Construction of topographical presentation used in comprehensive 2D-GC.

IV. PRINCIPLES OF TOFMS
In the most common form of TOFMS, ions created in the vacuum of an ion source
are accelerated by an electric field and allowed to drift through an evacuated
field-free region (flight tube) where they separate into groups (isomass packets)
according to their mass-to-charge ratio, as shown in Fig. 5. The time required
for an ion extracted from the source to reach the detector is measured and used
to calculate mass (21).
Ions gain energy when accelerated by an electric field in relation to their
charge, ze, so the classic equation describing the relationship between kinetic
energy eV, mass m and velocity v can be applied. Thus,
1
zeV mv 2 (1)
2
Over a fixed distance, the velocity is proportional to the reciprocal of the flight
time. Upon rearrangement, the relationship between flight time T and m/z is

m
Tk (2)
z
where k is the collection of constants. Assuming the number of charges z to be
1, which is predominantly the case with positive ions created by electron ioniza-
tion, this first-order relationship between time and m provides for a simple and
rapid calibration of the TOF mass spectrometer. In the rare event of ions being
formed with more than one charge, the relationship is still valid. However, the
measured times will be shorter, causing ions of higher masses and greater charge
to appear in the region of the spectrum where smaller masses having a single
charge appear.
In modern GC-TOFMS, computer assistance is mandatory, and the time-
of-flight measurements are made in the following manner. The computer sends
FIGURE 5 Mass analysis by TOFMS.

a signal to the mass spectrometer to begin the timing cycle. This triggers the
extraction of the ions from the source and initiates the timing sequence for all
of them. The arrival of the individual isomass ion packets at the detector deter-
mines the end of the flight time for each specific m/z. In reality, there is a fixed
delay between the time the computer initiates the extraction and when it actually
occurs, as well as between the time ions strike the detector and when they are
recorded. Fortunately, these delays are very short and reproducible and can be
accounted for in the calibration procedure by the addition of a single constant,
toffset. Hence, for all ions

m
Tmeasured toffset k (3)
z
The resolving power of TOFMS is challenged by the fact that because the
ionization occurs in the gas phase (3-dimensional space) all of the ions do not
begin from the same position and thus do not all end up with exactly the same
energy. These space and energy variations must be corrected for TOFMS to be-
come a viable technique.
The seminal work by Wiley and McLaren (22) led to the first commercial
TOF mass spectrometer. Using two electric fields, a small one for ion extraction
and a large one for ion acceleration, they were able to correct for the space and
energy variations and their two-field unit became the standard for TOFMS. How-
ever, despite the fact that the first experimental GC-MS system used a TOF mass
spectrometer (23), commercial application of this system to GC was not very
successful. This was due in a large part to the method of data collection. Only
a small part of the spectrum could be collected after each source extraction, using
a method called time slice detection (TSD). By gradually increasing the interval
between the moment of extraction and a narrow data collection window, or slice-
in-time, a spectrum was generated in a manner similar to that of a scanning instru-
ment. This made the system slow in spectral generation rate, and, due to the
increased pressure in the ion source from the GC carrier gas stream, it also suf-
fered from poor sensitivity and poor resolution. For these reasons, the technique
of GC-TOFMS was not further developed at that time.
With the advent and proliferation of high-resolution capillary GC, the ever-
increasing quest for higher resolution on the time axis led to a reevaluation of
the potential capabilities of TOFMS (9). Ion extraction rates in TOFMS typically
range from 5000 to 10000 per second, so it was apparent that the potential for
speed was immense because information sufficient for a mass spectrum was actu-
ally embedded in each transient waveform striking the detector. Indeed, the ad-
vances in microelectronics and digital computers enabled the development of
time-array detection (TAD) using an integrating transient recorder (ITR) in the
1980s (9,2426). In TAD, all of the ions throughout the entire m/z range are
measured for each transient. Because it is not convenient, or necessary, to convert

every transient into an individual mass spectrum, successive transients are
summed in a time-lock registry by the ITR to produce spectra at desired rates
from one per second to several hundred per second. Several advantages are gained
by using TAD, including (a) speedspectra may be generated at rates compatible
with requirements of the chromatographic resolution; (b) sensitivitysummation
of successive transients increases the magnitude of all ion measurements in the
spectrum; (c) signal/noisethe increase in the signal is complemented by the
reduction in noise, which is equal to the square root of the number of transients
summed to create each spectrum; (d) limit of detectionthe noise level is often
the major determinant in this measurement and it can be attenuated by increased
summations.
The construction and functioning of a modern GC-TOF mass spectrometer
with TAD is shown in Fig. 6. The molecules exiting from the GC are transferred
into the vacuum of the ion source, ionized by collision with an energetic focused
electron beam, and extracted from the source 5000 times a second by application
of a single high electric field. This high extraction field removes the detrimental
FIGURE 6 Modern Reflectron GC-TOF Mass Spectrometer System.

effects that the carrier stream pressure exerted on the earlier two-field source and
is possible because the ions are directed through a field-free region to an ion
mirror, which reflects them toward the detector (2729). The mirror corrects
for the spatial and energetic variations encountered in the three-dimensional ion
volume of the source. Because all of the ions experience the field of the mirror,
the simple relationship between total flight time and m/z is still maintained. After
the ion beam strikes the detector, the transient signals created are fed to the ITR
for digitalization, summation, and transfer to the computer for storage, future
processing, and output.
V. ANALYSES BY GCTOFMS
A. Data Collection
In relation to the arts and practices of modern chromatography, the ability for
rapid data collection presented by TOFMS can effectively be employed anywhere
between two extremes: (a) to obtain much more information using the conven-
tional chromatographic time frames or (b) to obtain the same information in much
shorter time frames. In all situations, the quantity of the data collected is greatly
increased over standard GC-MS. Where typical separation times are employed,
the spectral generation rates of 50 or more per second create data files with resolu-
tions on the chromatographic ordinate previously not attainable. Where coelu-
tions occur, this data density significantly enhances the deconvolution processes,
both in accuracy and speed. At the other extreme, where fast or time-compressed
chromatographic techniques are employed, spectral generation rates of 200 or
more per second produce resolutions along the time axis that were previously
unattainable. Indeed, in the mix and match between chromatographic separation
time and spectral generation rate, both of these analytical determinant axes can
simultaneously be enhanced over conventional GC-MS.
B. Data Analysis
Over the years, many approaches to the analysis of GC-MS data have been pro-
posed utilizing various algorithms, many of which are quite sophisticated, in
efforts to detect, identify, and quantitate all of the chromatographic peaks. Be-
cause the success or failure of these routines fundamentally depends on the qual-
ity of the data submitted to them, the results using data from scanning instruments
are often limited. Interestingly, a number of these algorithms perform well when
TOFMS data are submitted for analysis. The high density and quality of the data
collected by TOFMS enable accurate execution of the most powerful deconvolu-
tion routines available. For the exclusive processing of TOFMS data, the high
resolution of the time axis and the total lack of skewing permit the following
assumptions to be applied, enhancing both the speed and accuracy of the analysis
and greatly simplifying the complexity of the computational routines themselves.
1. All ions arising from a single compound exhibit a constant, concentra-
tion-independent, relative relationship with each other.
2. All ions arising from the same compound follow an identical chromato-
graphic dynamic in time.
3. Multiple components simultaneously present in the ion source do not
affect the fragmentation patterns of each other, and the measured ion intensities
are the linear sums of the ion intensities of the compounds present.
Data analysis generally involves a sequence of five steps. Typically, the
routines to perform these steps employ interactive operator direction with subse-
quent automatic execution. In some cases, however, complete analysis can be
predetermined and executed throughout with a minimum of operator interaction.
Indeed, the power of TOFMS with TAD can be readily appreciated by the speed,
accuracy, and ease with which the deconvolution routines based on the assump-
tions given above can be automatically executed. To accomplish these manually
would be time consuming and laborious, even with the assistance of computerized
graphics and interactive routines. Obviously, any logic applied automatically by
computer routines can be duplicated by human interaction using separately ad-
dressable functions in a significantly longer time frame.
1. Peak Finding
The first step in the analysis is to detect and locate every peak in each m/z channel.
Various algorithms can be used to differentiate a peak from noise in an ion chro-
matogram. Their logic usually defines an intensity threshold, which is some mul-
tiple over the background intensity, and must be exceeded for a specified mini-
mum time in order to avoid the registering of noise spikes. Next, the time of the
maximum intensity response in this region or, where necessary, the time of the
centroid center of the maximum response, is recorded. A peak position plot is
then generated along the time axis by adding an increment to the ordinate at each
time position in all of the ion chromatograms having a peak at that time. Because
all of the ions arising from the same compound will peak at the same point in
time, the ordinate will receive several increments at this point, making these plots
extremely effective in peak detection. Figure 7 illustrates the nature of the peak
position plot and shows the advantage of finding peaks by the automated system.
A visual inspection of the reconstructed total ion chromatogram (RTIC), upper
chromatogram, indicates the presence of nine components over a 30-second por-
tion of the elution. The vertical lines on the lower chromatogram constitute the
computer-generated peak position plot and confirm the presence of 24 compo-
nents over the same period. The accuracy of the determination of the presence
FIGURE 7 Determination of the presence and location of eluting peaks from Red Deli-
cious apple volatiles. The data were acquired at 30 spectra/sec using the LECO Pegasus
II TOFMS system. [Data courtesy of Dr. Randolph M. Beaudry, Michigan State University
Department of Horticulture (17).]
and location of every eluting peak is essential to the success of the following
steps in the GC-MS analysis.
2. Determination of Individual-Component Elution Profile

The second step is to determine the elution profile of each component in the
sample mixture. All of the ions arising from the same compound will exhibit the
same profile, so it is necessary to define only one ion profile for each component.
Thus, any m/z that is unique (not interfered with by ions of the same m/z that
arise from a different coeluting component) in the time period of an individual
elution can define the elution behavior of its originating component. Because of
the absence of skewing, the temporal behavior of all other ions arising from this
component will be the same. Figure 8a illustrates the resolution of a coeluting
peak when both components exhibit ions of unique m/z. By this means, determin-
ing the profile of any component with a unique ion is straightforward, presenting
a significant time saving at this stage. Moreover, as show in Fig. 8b, the presence
of at least one unique m/z for coeluting components allows simple mathematical
deconvolution to be applied, generating the profile of the component that does
not have a unique ion.
In situations where there is not a unique m/z to profile a specific component,
various computational routines can be employed that will identify a m/z that is
most nearly unique and create a profile of its temporal behavior. Examining the
ion chromatograms for maximum differentiation between intensities at peak cen-
ter, peak base, and background identifies candidates for the most unique m/z. In
these situations, the standard deconvolution routines of classic single-channel
chromatography can be employed to complete the assessment of the elution be-
havior and most closely approximate the true profile. The results of these pro-
cesses yield an elution function for each detected component, which can be used
for quantitation and can assist in the subsequent deconvolution processes.
3. Determination of the Correct Fragmentation Pattern (Mass

Spectrum)
After background subtraction, every spectrum across a chromatographically iso-
lated peak should be nearly identical and a true representation of that compound,
FIGURE 8 Determining the profiles of coeluting analytes where (a) each analyte contains
unique ions and (b) where one analyte exhibits only shared ions.
as shown in Fig. 9. Because the mass spectra for most compounds vary slightly
depending on the type of mass spectrometer the compound is analyzed on, for
convenience in this text, we will refer to those produced by TOFMS that are
consistently reproducible and concentration-independent to be accurate, or true,
mass spectra. In the case of coeluting, but not congruent, components, assignment
of unique ions and their relative intensities on the basis of their identical peak
position in time to the true mass spectrum of an originating compound is straight-
forward. Additionally, any nonoverlapping edge yields the true mass spectrum
of that temporarily isolated component. This information, coupled with a known
elution profile, permits simple spectrum subtraction to deconvolute coeluting
components. In these situations, as shown in Fig. 10, where partially coeluting
components each have at least one unique ion, simple edge subtraction guided
FIGURE 9 Uniformity and concentration independence of spectra generated using

TOFMS with TAD.
FIGURE 10 Deconvolution by edge stripping.
by symmetry positions on the known unique ion profile yields the complete true
mass spectrum of each component. Here the presence of a unique m/z for only
one of coeluting components often is sufficient for determining the true mass
spectrum for each component.
However, the presence of non-unique ionsi.e., ions in the elution window
of the same m/z but arising from different compoundsmake determination of
the true mass spectrum of an individual component more difficult. With a greater
number of coeluting components, the probability in any elution window of ions
having the same m/z but arising from different compounds is significantly in-
creased. These are called shared ions because, unlike the unique ions whose inten-
sity can be directly used, the measured intensity of the m/z must be allocated
among two or more components. Allocation of this sharing is critical to the rela-
tive abundance of the m/zs in each mass spectrum subsequently determined. In
these situations, because of the nonskewed nature of the data, each mass spectrum
produced by TOFMS will be the linear sum of the true mass spectra of the compo-
nents present, with each factored by its concentration in the ion source at the
moment of extraction. Consequently, simple equations can be used to deconvo-
lute or properly assign portions of the various shared ions to the different species
present in each acquired spectrum, even in situations where peak temporal separa-
tions are as small as 8 msec. This enables rapid and efficient determination of
the true spectrum for each component, as shown in Figure 11. The mass spectra
determined in this step can be submitted directly to library search routines for
qualitative analysis or be used in identifying analytical determinates for quantita-
tive analysis.
4. Qualitative Analysis
Typically, the determined fragmentation pattern (mass spectrum) of each as-
sumed compound is submitted for what is called a forward library search. In this
approach, the library is searched for entries that most nearly fit the pattern of the
candidate. There are several advantages of library searching. It is convenient,
fast, objective, and fairly comprehensive. It is nearly mandatory in situations
where the analyst doesnt have access to accepted spectral patterns or a series
of standard compounds. In addition, an answer is always provided, and several
candidate compounds may be listed as potential matches to the unknown spec-
trum. The candidates are listed in the order of confidence that the search algorithm
determines as corresponding to the best fit, or most probable correct answer.
There are several libraries and library search routines available that can be used
once their file and encoding formats are known. The library being used to provide
examples in this chapter is the NIST library. For this search routine, the confi-
dence is related to a similarity factor and is based on a mathematical calculation
incorporating the match of the m/zs present in the spectrum submitted and their
relative intensity interrelationships, with those of the library spectra. A perfect
match results in a similarity factor of 1000, whereas spectra with no m/zs in
common result in a value of 0. As a general guide, 900 or greater is an excellent
match, 800900 a good match, and 700800 a fair match. Less than 600 is a
poor match. Unknown spectra exhibiting a more complex fragmentation pattern
will tend to yield lower match factors than those exhibiting less complex fragmen-
tation. Additionally, components belonging to groups containing large numbers
of similar compounds will score lower than components belonging to groups
having smaller numbers of similar compounds. Often, knowledge of the sample
source, its possible chemical nature, and other factors need be considered in an
interactive dialogue between the computer and the analyst in achieving the correct
identification. Search routines are wonderful, but they are not omnipotent. There
is still room for human error! Figure 12 illustrates the accuracy of deconvolution
for two coeluting apple flavor components (17). The RTIC indicates the presence
of a single 2-second wide peak at t115.3 sec, and the mass spectrum at this
point represents the combination of both analytes. Deconvolution provides iso-
lated spectra that match closely to the NIST reference spectra.
FIGURE 11 Deconvolution of limonene and eucalyptol in a spearmint flavor mixture us-

ing the LECO Pegasus II acquiring at 30 spectra/sec. Initial examination of the RTIC
indicates the presence of only limonene at t 141 seconds, whereas deconvolution reveals
a minor amount of eucalyptol as well. (Data courtesy of Dr. Rajesh Pandya, Robertet
Flavors, Inc.)
FIGURE 12 Typical qualitative results using the NIST Library.

5. Quantitative Analysis
Once a compound has been identified and its elution profile determined, the final
(and optional) step involves the selection of an m/z having sufficient intensity
and uniqueness for quantitation of that compound. The area under the elution
profile can then be used as the quantitative determinant. In situations where sensi-
tivity is critical, more than one m/z, unique or corrected to uniqueness, can be
summed to gain an aggregate profile that then can be used as the determinant.
Once the analytical determinant has been identified, it can directly be used for
relative determinations of concentrations but must be calibrated for more accurate
measurements.
Calibration of the quantitative determinant usually involves the measure-
ment of various known concentrations of the compound under similar instrumen-
tal conditions and in a similar sample milieu. Plotting the concentration versus
the area under the elution profile of the determinant creates a working curve.
The area measured for subsequent analyses of the same compound at unknown
concentrations are referenced to this working curve for assessment. The working
curve is the most accurate method of calibration and permits measurement of
nonlinear relationships. Because of the complexity of GC-MS systems, determi-
nant areas can be ratioed against a constant internal standard, often co-injected,
prior to their utility in establishing and referencing a working curve.
Figure 13 shows a working curve with an extended working range that
FIGURE 13 Extended dynamic range for limonene using separate m/zs (30). The use of
both m/z 68 and m/z 136 for quantitation provides a dynamic range of 0.5 pg to 500 ng.
capitalizes on the uniformity of the fragmentation relationships found in array

detection (30). The portion of the working curve relating to the higher concentra-
tions employs a minor fragment ion as the analytical determinant and the portion
relating the range of the lower concentrations uses a major ion fragment as the
determinant. This combination results in the measurement of limonene over a 6
order-of-magnitude concentration range, from 0.5 pg to 500 ng.
Internal standards can also be used directly for quantitation without the
creation of a working curve. This approach is less accurate than full calibration
but is convenient under conditions where exact measure is not required. In this
case, the response area of a single concentration, usually in the mid-range of
the expected concentrations, is measured and its area ratioed against the internal
standard to obtain an analytical factor that can then be used to convert future
measured ratios into the desired concentration units, assuming a linearity of in-
strument response.
Typically, all of the above routines employ interactive operator direction
with subsequent execution. However, complete analysis routines are often prede-
signed and automatically executed to accommodate automatic sampling systems
and batch analyses.
C. Applications of GC-TOFMS
1. Analysis of Complex Mixtures
While there are limitless opportunities to use TOFMS in conventional GC-MS
analyses, there are also a variety of applications developed using GC-TOFMS
that take advantage of the potential to provide more information in less time.
Figure 14 illustrates the 3-minute analysis of a complex citrus standard mixture,
using time-compressed chromatography, which normally requires 4060 minutes
using conventional GC-MS. Compression of the time axis occurs without any
loss in analytical resolution. This time-compression is potentially useful for a
variety of high-throughput screening applications; and thus, although only 53
peaks are visible in the RTIC, deconvolution detected 198 analytes. Indeed, it is
this rapid and comprehensive qualitative analysis capacity that documents one
of the most powerful aspects of TOMS with TAD.
2. Fast Chromatography
Another class of applications involves the use of fast chromatography, where the
chromatographic technology applies sample compression techniques to generate
narrow peak widths in addition to compressing the analysis time, as described
earlier. Figure 15 illustrates the analysis of volatile contaminants (17). A cryofo-
cusing gas injector system (Chromatofast, Inc.) is combined with a HP6890 gas
chromatograph and the LECO Pegasus II TOF mass spectrometer to generate
FIGURE 14 Analysis of a complex citrus standard mixture. Although only 53 peaks are
evident in the RTIC, deconvolution revealed the presence of 198 compounds. (Data cour-
tesy of Dr. Nancy Myers, LECO Corporation.)
peak widths of only a few hundred milliseconds, for analytes also separated by
only a few hundred milliseconds. Deconvolution generates individual mass spec-
tra for each component. The use of fast chromatography in this application re-
duces the analysis time and also lowers the achievable detection limit by generat-
ing narrow, and thus taller, peaks. This particular application is illustrative of
throughput gains that can be achieved with analyses that require rapid feedback
or nearly constant monitoring of dynamic systems, and applications to flavor and
fragrance analyses. Thus, significant reductions in analysis time must also be
accompanied by relatively short data processing times. The quality of the mass
spectral data generated using TOFMS combined with efficient deconvolution al-
gorithms, result in data processing times of only a few seconds for the above
application.
3. Trace Analyses
Many quality control analyses require the detection of minor contaminants in a
sample product. These types of analyses can be inherently difficult because, by
definition, the contaminants being searched for will occur in low amounts, and
the number and type of contaminants present in the sample may be unknown. The
analysis is further complicated if the contaminant coelutes with a more dominant
FIGURE 15 Fast chromatography of volatile contaminants. Data were acquired using fast
chromatography and TOFMS at 50 spectra/sec. (From Ref. 32).
FIGURE 16 Trace contaminant analysis. Deconvolution and subsequent analysis of con-

taminant standards confirmed the identity of each analyte. (Data courtesy of Dr. Nancy
Myers, LECO Corporation.)
product analyte. Figure 16 illustrates an extreme example of this application type,

where several contaminants elute under the solvent peak of a flavor sample. In
this case, the contaminants were 400,000 times less abundant than the propylene
glycol solvent, and yet subsequent deconvolution produced library-matchable
spectra for each one.
4. Comprehensive 2D-GC/TOFMS
As an alternative method for the analysis of very complex mixtures, comprehen-
sive 2D-GC combines two chromatographic retention characteristics to distin-
guish between similar analytes. This technique has been successfully applied in
the analysis of petroleum samples where many hundreds of analyte species are
present. Yet, while the benefit of analyte separation is realized in some cases,
identification of many analyte peaks has thus far been precluded. Figure 17 illus-
trates the benefit of multichannel detection in the analysis of herbicides (19). The
comprehensive 2-D field was generated using a 3 m, 0.25 mm I.D., 0.25 m film
thickness DB-5 first column and a 1.5 m, 0.1 mm I.D. 0.1 m film thickness
OV-1701 second column. The custom-designed modulator operated at a period
of 4 seconds. The LECO Pegasus II TOF mass spectrometer was used as the
detector, operating at 100 spectra/sec. The topographical chromatogram gener-
ated in this application uses two time axes and exhibits a complex background
from which the analytes must be discriminated. The combination of 2D-GC with
MS creates a fourth dimension of information to aid in definitive identification.
As shown in the deconvoluted chromatogram of the first peak cluster, the modula-
tion process generates very narrow peak widths of approximately 125 msec with
each compound analytically resolved. These narrow peak widths mandate high-
speed detection that, combined with the need for analyte identification, generally
precludes all detection methods other than TOFMS as the method of choice.
5. Deconvolution by Differential Fragmentation

To date, deconvolution routines have realized their differentiating parameters
using intervals of time, because rarely do components elute having both identical
ion fragmentation patterns and identical chromatographic temporal behavior.
Time-based deconvolution routines struggle, however, when compounds generat-
ing the same fragmentation pattern partially coelute. In these cases, other more
sophisticated routines must be employed to determine the individual elution man-
ifolds, and they completely fail when compounds having the same fragmentation
pattern exactly coelute in time (congruent coelution). In these situations, the pa-
rameters of the chromatographic separation (column type and resolving power,
carrier gas, temperature program, etc.) are changed in attempts to achieve even
marginal temporal separation. Unfortunately, extensive method development is
time consuming and costly, and some chemical isomers are unresolvable even
under optimal conditions.
An unforeseen advantage of TOFMS with TAD has been manifested in the
reproducibility of the ion fragmentation patterns. This presents the m/z axis itself
to be used in deconvolution. To date, information on this axis has been used only
to calculate confidence factors for the presence of components detected by the
time-based routines and not to deconvolute coeluting species. Because of the
unskewed quality of the TOFMS data, coeluting species, usually isomers, gener-
ating all ions of the same m/z but in differing relative intensities, can be deconvo-
luted by assigning appropriate amounts of the measured m/z intensities to each
FIGURE 17 2D-GC-TOFMS of a herbicide mixture.

species. This involves a search for differential fragmentation between each spe-
cies that can be used to determine the relative amounts of each using simple
linear equations.
Figure 18 illustrates the method of deconvolution by differential fragmenta-
tion applied to 2,5- and 2,6-dimethylpyrazine, a pair of isomers found in peanut
butter that coelute under common chromatographic conditions (31). In this case,
the relative abundance of the ions at m/z 39 and m/z 42 is different for each
isomer. Because of the reproducibility of the fragmentation pattern, this rather
small differentiation can be used to calibrate a deconvolution scale that allows
changes in the relative abundance of these two m/zs to be used for quantitation.
Once the ratio of these m/zs versus relative concentration is calibrated, a single
mass spectrum measured at the apex of the chromatographic peak will yield the
relative abundance of each coeluent. Quantitation is finalized by integration of
FIGURE 18 Deconvolution by differential fragmentation of geometric isomers. Analysis

of a standard mixture of 80% 2,5- and 20% 2,6-dimethylpyrazine, using the working curve,
extrapolated to 75% 2,5-dimethylpyrazine.
the coeluting peak along the time axis with subsequent apportionment of this
sum into the two components on the basis of this measured ratio.
In an interesting manner, the technique of differential fragmentation closes
the loop on the application of deconvolution. Where small temporal differences
exist between elution peaks, success has been shown in situations in which each
analyte has a unique ion; in which only one coeluting analyte has a unique ion;
and in which there are no unique ions. This technique applies where there is
complete coelution, no temporal differentiation, and all the ions are shared. It
cant get any more complex than that.
REFERENCES
1. J. M. Davis and J. C. Giddings, Statistical theory of component overlap in multicom-

ponent chromatograms, Anal. Chem. 55:418 (1983).
2. J. T. Watson, Introduction to Mass Spectrometry, 3rd Ed., Lippincott-Raven, Phila-
delphia, 1997.
3. O. D. Sparkman, Mass Spec Desk Reference, Global View, Pittsburgh, 2000.
4. National Institutes of Standards and Technology, NIST/EPA/NIH Mass Spectral
Library, United States Department of Commerce.
5. Wiley Registry of Mass Spectral Data, Palisade Corporation, Newfield, NY.
6. C. G. Enke, J. T. Watson, J. Allison, and J. F. Holland, Analysis of data rate require-
ments for high performance GC/MS: the case for TOFMS with time-array detection,
Proc. 37th ASMS Conf. Mass Spectrometry and Allied Topics, Miami Beach, Flor-
ida, May 2126, 1989.
7. S. Chesler and S. P. Cram, Effect of peak sensing and random noise on the precision
and accuracy of statistical moment analyses from digital chromatographic data, Anal.
Chem. 43:1922 (1971).
8. B. D. Gardner and J. F. Holland, High speed, high throughput GC-MS using time-
of-flight mass spectrometry with time array detection, Proc. of 44th ASMS Conf.
On Mass Spectrometry and Allied Topics, Portland, OR, May 1216, 1996.
9. J. F. Holland, C. G. Enke, J. Allison, J. T. Stults, J. D. Pinkston, B. Newcome,
and J. T. Watson, Mass spectrometry on the chromatographic time scale: realistic
expectations, Anal. Chem. 55:997A, (1983).
10. B. D. Gardner, Dissertation, Michigan State University, 1997.
11. R. Grix, R. E. Tecklenburg, Jr., J. T. Watson, J. F. Holland, G. E. Yefchak, B. D.
Gardner, J. Allison, C. G. Enke, and H. Wollnik, A time-of-flight mass spectrometer
for high-speed Chromatography, Proc. of the 39th ASMS Conf. Mass Spectrometry
and Allied Topics, Nashville, TN, May 1924, 1991.
12. J. F. Holland and B. D. Gardner, Enhancement of analytical resolution with high
speed GC/TOFMS, Proc. 45th ASMS Conf. Mass Spectrometry and Allied Topics,
Palm Springs, CA, June 15, 1997.
13. B. D. Gardner and J. F. Holland, High speed, high throughput GC-MS using time-
of-flight mass Spectrometry with time array detection, in Proc. 44th ASMS Conf.
Mass Spectrometry and Allied Topics, Portland, OR, May 1216, 1996.
14. C. C. Sweeley, W. H. Elliott, I. Fries, and R. Ryhage, Mass spectrometric determina-
tion of unresolved components in gas chromatography effluents, Anal. Chem. 38
(1966).
15. J. F. Holland, C. C. Sweeley, R. E. Thrush, R. E. Teets, and M. A. Bieber, On-line
computer controlled multiple ion detection in combined gas chromatography-mass
spectrometry, Anal. Chem. 45:2 (1973).
16. C. G. Enke, J. F. Holland, R. D. McLane, and G. E. Yefchak, Time-compressed
chromatography in mass spectrometry, United States Patent 5,175,430, 1992.
17. J. Song, B. D. Gardner, J. F. Holland, and R. M. Beaudry, Rapid analysis of volatile
flavor compounds in apple fruit using SPME and GC/time-of-flight mass spectrome-
try, J. Agric. Food Chem. 45:1801 (1997).
18. Z. Liu and J. B. Phillips, Comprehensive two-dimensional gas chromatography using
an on-column thermal modulator interface, J. Chromatogr. Sci. 29:227 (1991).
19. B. D. Gardner, J. F. Holland, D. G. Patterson, Jr., and J. Grainger, Rapid environmen-
tal toxicant screening using multi-dimensional GC-time-of-flight mass spectrometry,
presented at 23rd Annual Conf. Fed. Analytical Chemistry Spectroscopy Societies,
Kansas City, MO, Sept. 29Oct. 4, 1996.
20. M. Van Deursen, J. Beens, J. Reijanga, P. Lipman, C. Cramers, and J. Blomberg,
Group-type identification of oil samples using comprehensive two-dimensional gas
chromatography coupled to a time-of-flight mass spectrometer, J. High Resolution
Chromatography 23:7 (2000).
21. See for example, R. J. Cotter, Time-of-flight mass spectrometry, American Chemical
Society, 1997.
22. W. C. Wiley and I. H. McLaren, Time-of-flight mass spectrometer with improved
resolution, Rev. Sci. Instrum. 26:1150 (1955).
23. R. Gohlke, Time-of-flight mass spectrometry and gas-liquid partition chromatogra-
phy, Anal. Chem. 31:535 (1959).
24. J. F. Holland, B. N. Newcome, R. E. Tecklenburg, M. Davenport, J. Allison, J. T.
Watson, and C. G. Enke, Design, construction, and evaluation of an integrating tran-
sient recorder for data acquisition in capillary gas chromatography/time-of-flight
mass spectrometry, Rev. Sci. Instrum. 62:69 (1991).
25. E. D. Erickson, C. G. Enke, J. F. Holland, and J. T. Watson, Application of time
array detection to capillary column gas chromatography/conventional time-of-flight
mass spectrometry, Anal. Chem. 62:1079 (1990).
26. J. F. Holland, J. Allison, J. T. Watson, and C. G. Enke, Achieving the Maximum
Characterizing Power for Chromatographic Detection by Mass Spectrometry, in
Time-of-Flight Mass Spectrometry (R. J. Cotter, ed.), American Chemical Society,
1994.
27. V. I. Karataev, B. A. Mamyrin, and D. V. Shmikk, Sov. Phys. Tech. Phys. 16:1177
(1972).
28. B. A. Mamyrin, V. I. Karataev, D. V. Shmikk, and V. A. Zagulin, Sov. Phys. JETP
37:45 (1973).
29. R. Grix, R. Kutscher, G. Li, U. Gruner, and H. Wollnik, Rapid Comm. Mass Spec-
trom. 2:83 (1988).
30. V. Artaev, K. McNitt, M. F. Merrick, and J. Mitchell, The extension of instrument

dynamic range in GC/TOFMS by simultaneous sampling, Book of Abstracts, Pittcon
98, New Orleans, LA, March 15, 1998.
31. B. D. Gardner and J. F. Holland, Fast mass spectrometer detection for fast GC:
shifting the rate limiting step from separations to sampling, Book of Abstracts, Pitt-
con 98, New Orleans, LA, March 15, 1998.
32. B. D. Gardner, R. E. Berkley, Jr., and J. F. Holland, Rapid analysis of TO-14, air-
borne contaminants by fast GC/TOFMS, Proc. 44th ASMS Conf. on Mass Spec-
trometry and Allied Topics, Portland, OR, May 1216, 1996.
6
Modern Methods for Isolating
and Quantifying Volatile Flavor
and Fragrance Compounds
Peter Werkhoff, Stefan Brennecke, Wilfried Bretschneider,

and Heinz-Jurgen Bertram
Haarmann & Reimer GmbH, Holzminden, Germany
I. INTRODUCTION
The characteristic flavors of food and beverages as well as the pleasing scent of
perfumery products are generally the result of extremely complex multisubstance
mixtures, containing several hundred compounds with different chemical struc-
tures. So in most cases, the flavor impression or scent character that we perceive
as a single sensation when we enjoy one of these products is a complex sensory
impression of many individual substances in specific concentration ratio. These
volatile components, which are present in only minor amounts (ppm to ppt range),
decisively influence the enjoyment and the acceptance of foodstuffs, perfumes,
and personal care or household products. Only in rare cases are individual compo-
nents (flavor impact compounds) responsible for odor and taste.
The purpose of modern flavor analysis is to qualitatively and quantitatively
decipher the flavor profile defined by nature and, most important, to recognize
and prioritize organoleptically interesting flavor compounds and distinguish them
from other volatile compounds that may have no organoleptic relevance or may
merely balance a flavor. The methods used today are liquid and gas chromatogra-
phy for separation, as well as combinations of chromatographic and spectrometric
methods for identification. Organoleptically active constituents of a flavor com-
plex are detected by a combination of gas chromatography/olfactometry (GC/O),
139
140 Werkhoff et al.
and their quantitative contribution is determined by aroma extract dilution analy-

sis (AEDA), CHARM analysis, or the OSME concept.
But identification and the search for organoleptically relevant key compo-
nents are among the last steps in the analytical procedure used to characterize
flavors. Modern flavor analysis always begins with sample preparationi.e., the
isolation and separation of flavor components from natural productsand the
low concentration at which typical flavor compounds exist in foods usually makes
this a laborious task. This critical step in flavor analysis can involve a very wide
variety of methods and is at least as important as the subsequent operations of
separating the individual components and identifying their structures. The com-
ponents being analyzed, which are often highly labile or reactive, should undergo
as little discrimination or modification as possible.
Even today, however, with all our technical progress, there is unfortunately
still no universal isolation method capable of covering the entire spectrum of
volatile flavor compounds. For this reason, depending on the quality and character
of the particular natural product or food, many different methods and instrumental
systems are used to isolate and concentrate volatile flavor components. These
very diverse sample preparation techniques can, however, influence both the qual-
itative and quantitative composition of the flavor compounds and thus, of course,
the organoleptic properties of the total flavor concentrates. The methods used to
isolate volatile flavor compounds are therefore always very important, and it must
never be forgotten that even the most powerful analytical system cannot compen-
sate for errors made while recovering a flavor concentrate. The composition of
a flavor isolateand its organoleptic propertiesis thus crucially dependent on
the sample preparation method that is used. The most important task in the choice
of the isolation procedure is to ensure that the total flavor extract faithfully repli-
cates the characteristic and typically olfactory and gustatory features of the ana-
lyzed productbecause only a representative flavor extract can serve as the basis
for optimum and reliable flavor analysis.
Well-established isolation techniques have been available for many years,
and they still remain valid today:
Liquid/liquid extraction (Kutscher-Steudel, rotational perforation)

Liquid/solid extraction (Soxhlet principle)
Solid-phase extraction (SPE)
Likens-Nickerson simultaneous distillation/extraction (including in vac-
uum)
Simultaneous distillation/adsorption
Microwave extraction
Accelerated solvent extraction (ASE)
Headspace methods (static headspace method, dynamic headspace method
followed by thermal or solvent desorption)
Isolating and Quantifying Volatile Compounds 141
Headspace technology including Zenith Trap collector

Vacuum headspace technique
Closed-loop stripping analysis (CLSA)
High-vacuum sublimation according to Prof. Grosch
Gas extraction
Reflux trapping according to Prof. Jennings
Concentration of headspace components from aqueous/ethanolic samples
according to Prof. Rapp
Extraction with liquid CO2 according to Prof. Jennings
Evaporation through a membrane (pervaporation)
Direct thermal desorption
These established methods for isolating and concentrating flavor compounds, and
their effects on the composition of the aroma concentrates, are described in com-
prehensive detail in the literature (127).
This chapter discusses four additional sample preparation methods that
have been successfully used for some time in our laboratory in the field of flavor
and fragrance research.
II. SOLID-PHASE MICROEXTRACTION (SPME)
Solid-phase microextraction (SPME) was developed in 1988 by Prof. Janusz

Pawliszyn and his coworkers at Waterloo University in Ontario, Canada, and
commercialized in 1993 by Supelco (2832). Solid-phase microextraction is an
innovative, completely solvent-free sample preparation method that can be used
to concentrate flavors and fragrances and that significantly reduces the time and
cost required for sample preparation; it therefore represents an excellent alterna-
tive to conventional sample extraction techniques. The advantages of SPME are
particularly remarkable because this solvent-free sample preparation method does
not require any complex or expensive equipment. This highly effective technique
can be used to concentrate volatile organic compounds from liquid samples (im-
mersion SPME) or from the headspace above a liquid or solid sample (headspace
SPME).
The analytes from a liquid or from the headspace above a sample are con-
centrated on a fused silica fiber coated with a polymer film, causing an equilib-
rium distribution of the analytes to be established between the stationary phase
(the microfiber) and the aqueous or gas phase (the sample). Once equilibrium
has been established between sample and fiber, the concentrated compounds are
thermally desorbed in the injector of a gas chromatograph and transferred to the
capillary column (Fig. 1). The parameters to be optimized here are the nature of
the fiber coating (fiber polarity) and fiber film thickness, as well as the concentra-
142 Werkhoff et al.
FIGURE 1 Functional principle of the SPME method. (A) Configuration of the SPME
sampler; (B) sampling and extracting flavor compounds; (C) thermal desorption.
tion and desorption time and temperature. Desorption can be performed in much
the same way as a conventional sample injection, so the procedure is relatively
simple and quite reproducible. Cryofocusing of the flavor compounds during ther-
mal desorption is generally not necessary.
The design of a Supelco SPME holder and the sequence of extraction and
desorption steps are exhaustively documented in the literature (3335). All of
the theoretical and practical aspects of solid-phase microextraction are discussed
in detail by the inventor of the technique in his recently published monograph
(36).
At the moment, six different polar phase films are available for extracting
organic compounds from various sample matrices:
PDMS, Polydimethylsiloxane (7 m, 30 m, 100m)
PA, Polyacrylate (85 m)
PDMS/DVB, Polydimethylsiloxane/Divinylbenzene (65 m)
CW/DVB, Carbowax/Divinylbenzene (65 m, 70 m)
Carboxen-PDMS, Carboxen-Polydimethylsiloxane (75 m, 85 m)
DVB-Carboxen, Divinylbenzene/Carboxen (50/30 m)
To select the optimum fiber type for the analysis of flavor compounds, we used
a test mixture containing a variety of functional groups, and performed extrac-
tions from both the aqueous phase and the headspace. A comparison of the ad-
sorption capabilities of four different polar fiber types, using our test mixture, is
indicated in Figs. 2 and 3, which clearly show the differing absorption affinities
of the various flavors and fragrances. Figures 2 and 3 impressively reflect the
differences in selectivity of the individual fiber types and, at the same time, show
that the polydimethylsiloxane/divinylbenzene fiber represents an appropriate sta-
tionary phase for flavor and fragrance research.
The adsorption of organic compounds from aqueous samples is influenced,
among other factors, by the molecular size, polarity, and water solubility of a
substance. The recovery of difficult-to-extract components from aqueous samples
can also be improved by adding salt, by establishing the correct pH and fiber
immersion depth into the liquid phase, and by a certain amount of agitation during
sampling. The mechanical motion during concentration prevents the formation of
an analyte-depleted zone around the fiber; in other words, diffusion-controlled
transport of organic analytes into the fiber is promoted by agitation of the solution,
thus reducing the extraction time. The significant effect of adding sodium chloride
to the aqueous phase is illustrated in Fig. 4: The salting-out effect enhances the
extraction of many chemical compounds. This effect is particularly evident in
the case of headspace SPME analysis. Another important parameter for the head-
FIGURE 2 Determination of recovery rates for four different fiber types using immersion
SPME.
144 Werkhoff et al.
FIGURE 3 Determination of recovery rates for four different fiber types using headspace
SPME.
space SPME technique, which can markedly improve sensitivity and speed of
the analysis, is selection of the correct temperature.
Because the SPME method is characterized by a sufficiently linear range, it
can also be used successfully to quantify flavor compounds, although an accurate
quantitative determination of individual components requires considerable effort
in terms of calibration. The concentration of the analytes concentrated onto the
SPME fiber is proportional to the concentration in aqueous solution or in the gas
phase, so that quantification can be performed by attaining equilibrium, or by
always using the same fiber exposure times. The time required to establish equi-
librium depends on the distribution coefficients of the analytes and the film thick-
ness of the phase. Practical experience indicates that maintaining identical times
is more critical than establishing complete equilibrium. With this approach, even
kinetically less favorable substances can be precisely quantified even before the
equilibrium concentration has been reached. For routine analysis, therefore, it is
not absolutely necessary to reach complete adsorption equilibrium, provided the
fiber exposure time is always kept exactly constant.
The SPME method should be regarded as an alternative to the conventional
static or dynamic headspace technique, and at the same time serves as a substitute
FIGURE 4 Effect of the addition of sodium chloride on signal intensity (immersion

SPME, PDMS/DVB fiber).
for the extraction of organic compounds from aqueous samples using organic
solvents. As compared with other headspace techniques, the particular advantage
of headspace SPME is that it actually reflects the original composition of flavor
compounds in the gas phase above a producti.e. the olfactory impression per-
ceived by the nose is also in fact analytically detected.
The SPME technique (not to be confused with the SPE method) offers a
fast and easy way to identify flavor compounds in foods, to check flavor stability,
to detect off-flavor components and package-related contaminants, and so forth.
The literature has already reported on numerous applications in the field of envi-
ronmental, petrochemical, botanical, forensic, clinical and pharmaceutical analy-
sis (37). In addition, numerous flavor studies as well as food applications using
the SPME sampling technique have already been described (3760).
Figures 5 through 7 show a few applications from our laboratory, demon-
strating the importance of SPME for the extraction and analysis of flavor com-
pounds from a variety of foods. Figure 5 illustrates headspace SPME concentra-
tion of flavor compounds from several species of passion fruit (Passiflora spp.).
The various organoleptic properties of the three varieties of fruit are nicely repro-
duced in the resulting flavor profiles, i.e. the SPME technique can reliably be
used to characterize fruits and differentiate among varieties.
Figure 6 shows that solid-phase microextraction is suitable for analyzing
spice flavors as well. A headspace SPME sample of the volatile essential oil
146 Werkhoff et al.
FIGURE 5 Extraction of passion fruit flavor compounds using headspace SPME (PDMS/
DVB fiber).
components was taken using only 30 mg of Chinese mountain pepper. No other

sample preparation technique is capable of yielding such an intensive gas chro-
matogram so quickly and with such small quantities of sample.
Figure 7 illustrates the enormous advantage of the headspace SPME method
as compared with the static headspace technique, using the example of analytical
investigation of a dry apricot flavor. While the time expended is the same, the
FIGURE 6 Headspace SPME (PDMS/DVB fiber) flavor analysis of Chinese mountain

pepper.
improved sensitivity and higher extraction yield of the SPME method are clearly
evident.
Although SPME has been widely used for highly efficient extraction of
food components, little is known about the applicability of this technique for
monitoring fragrance materials in household cleaning products, soaps, detergents,
cosmetics, toiletries, etc. (6163). Until recently, costly and laborious sample
preparation techniques were required to isolate, separate, and identify fragrance
components in perfumery matrices.
Using immersion and headspace sampling, we show in Figs. 8 through 12
that the SPME method can also be used for fragrance analysis and is applicable
to a wide variety of sample types. With the aid of SPME technique, analytical
chemists will be freed from the complex and time-consuming classic sample
cleanup and preparation procedures that are currently used.
Figure 8 represents the analysis of an eau de toilette. A comparison of
direct injection versus immersion SPME is shown. The SPME chromatogram is
the result of a 5-minute room temperature SPME sampling of 15l EDT dissolved
in 1.5 ml distilled water. No significant difference in the fragrance pattern of
both chromatograms is evident.
The fragrance chemicals composing the scent of soap are usually complex
mixtures of alcohols, aldehydes, ketones, esters, and terpenoids. These compo-
nents are generally isolated from the soap matrix by conventional methods, such
148 Werkhoff et al.
FIGURE 7 Comparison of static headspace analysis with headspace SPME (PDMS/DVB

fiber): dry apricot flavor (20 m 0.18 mm I.D. DB WAX; 0.3 m film thickness).
as solvent extraction, steam distillation, or trapping of the volatiles on adsorbents

prior to gas chromatographic analysis. Few reports on headspace methods for
these components have been reported to date. We have applied headspace SPME
for the detection of fragrance materials in soap. In addition, the respective per-
fume oil was also analyzed by HS-SPME and was compared with perfume oil
volatiles directly injected onto the GC column. Obviously, all three chromato-
grams are rather similar in their fragrance profiles (Fig. 9). Only the most volatile
components are not adequately released from the soap matrix. Several reasons
could be responsible for this phenomenon: among other things, volatilization dur-
ing production and storage, stability problems, or interactions of fragrance com-
ponents with the soap matrix. Using conventional methods, a large amount of
sample is normally necessary. Contrary to this, the SPME method is advanta-
geous because it can extract a suitable amount from even a very small soap sample
(20 mg). Thus, in comparison with traditional procedures SPME is not so labori-
FIGURE 8 SPME profile of an eau de toilette using a PDMS/DVB fiber. Gas chromato-
grams of perfume oil components by direct injection (top), and by immersion-SPME
(bottom).
ous and material-consuming. Furthermore, the use of solvent is eliminated in the

extraction step and no concentration step is necessary. Moreover, when working
with large amounts of samples, high levels of background may have a negative
effect on the analysis and may complicate the interpretation of results.
Further applications of headspace SPME are illustrated in Fig. 10. The
chromatograms presented in Fig. 10 were obtained by extraction of small samples
of washing powder, fabric softener, and shampoo equilibrated in a vial at 50C
for 30 minutes and then extracted for 30 minutes using a 65-m PDMS/DVB
fiber. Once again, the advantages of SPME are that the technique is simpler and
faster compared with conventional extraction methods, and much cleaner ex-
tracts are obtained without disturbing or confusing background interferences.
There is a rapidly expanding literature on the application of SPME for the
analysis of essential oils and plant systems, and the utility of SPME for monitor-
150 Werkhoff et al.
FIGURE 9 SPME analysis of soap fragrance using a PDMS/DVB fiber. Chromatographic

comparison of perfume oil, HS-SPME of perfume oil, and HS-SPME of soap (20 m
0.18 mm I.D. DB-1; 0.4 m film thickness).
FIGURE 10 Gas chromatograms of extracted fragrance volatiles from washing powder,

fabric softener, and shampoo by HS-SPME, using the 65 m PDMS/DVB fiber.
FIGURE 11 Headspace SPME extraction of living flowers. Conditions: PDMS/DVB 65
m fiber, extraction for 30 minutes at room temperature. Rose (top), lavender (middle),
lily (bottom).
Isolating and Quantifying Volatile Compounds
FIGURE 12 SPME (PDMS/DVB fiber) in essential oil analysis. Direct injection of rose oil volatiles compared with the components obtained
by fiber extraction (TIC). Peak identities are as follows: 1, -pinene; 2, -pinene; 3, myrcene; 4, cis-rose oxide; 5, trans-rose oxide; 6,
153
linalool; 7, citronellol; 8, nerol; 9, geraniol; 10, 2-phenylethyl alcohol; 11, nonadecane; 12, (Z)-9-nonadecene; 13, methyl eugenol; 14,
eugenol; 15, farnesol.
154 Werkhoff et al.
ing the volatiles emanating from living plants is well documented in the literature.
SPME has been successfully applied for the extraction of volatile and semivola-
tile organic compounds from plant materials such as leaves, stems, and roots
(6470).
Figure 11 represents the volatiles obtained by SPME sampling from one
whole rose, lavender, and lily blossom. The SPME fiber was thermally desorbed
in a gas chromatograph coupled with mass spectrometry. The numbered compo-
nents were identified by comparison of their mass spectra and linear retention
indices with those of reference standards. To get as close as possible to the true
fragrance profile of a plant or flower material, direct headspace sampling or dy-
namic purge and trap headspace analysis are frequently applied. In comparison
with dynamic headspace sampling, however, SPME shows better precision and
accuracy, as well as higher sensitivity. As extraction and concentration from the
intact petals are combined, all of the analytes extracted are completely introduced
into the analytical system. At the same time, SPME offers a relatively simple,
yet rapid, inexpensive, solventless technique for the determination of the enantio-
meric composition of the optical isomers in natural flavors, essential oils, or head-
space samples of blossoms. For example, the potential of HS-SPME coupled with
a chiral column for characterizing linalool in a lily blossom is demonstrated in
Fig. 11 (bottom). Through the combination of nonchiral (30 m 0.25 mm I.D.
DB-WAX; 0.25 m d f) and chiral phase (25 m 0.25 mm I.D. 2,3-diethyl-6-
tert.butyldimethylsilyl--cyclodextrin; 0.25 m d f) using multidimensional gas
chromatography, knowledge of the distribution of enantiomers in plant materials
and essential oils has been greatly improved. R-()-linalool (93.0%) was found
to be the dominant enantiomer in the lily blossom. The somewhat uncommon
and strange scent of this lily blossom variety is due to a relatively high concentra-
tion of p-cresol, which creates a slightly fecal and horse stablelike sensory im-
pression.
Moreover, headspace SPME in combination with HRGC is an ideal and
extremely fast approach for characterizing and screening quality and composition
of essential oils. For example, in Fig. 12 the direct injection of a rose oil solution
in comparison with the SPME/TIC profile is shown. All constituents of the rose
oil including low and higher boiling components could be found by SPME. How-
ever, significantly different distributions of volatile components in both samples
are observed. It is evident that the SPME extract contains higher concentrations
of low boiling compounds such as -pinene, -pinene, myrcene, rose oxides, or
linalool. In contrast, the higher boiling and more polar components are better
represented in the oil chromatogram, e.g., citronellol, nerol, geraniol, methyl eu-
genol, eugenol and farnesol. Nonadecane and (Z)-9-nonadecene, for example,
are present in the oil but completely missing in the SPME profile. Therefore,
SPME is well suited for qualitative analysis of essential oil components, but it
has to be taken into consideration that the precise quantitative determination of
volatile components may differ from the respective oil composition. If a precise
quantitative measurement of essential oil components is desired, direct injection

of the oil dissolved in an organic solvent will provide optimal information.
It is apparent from these few examples that SPME is a modern alternative
to current sample preparation technology and a powerful tool for the determina-
tion of volatile components in fragrance research. Our studies have indicated
that solid-phase microextraction is far superior to most conventional extraction
methods for the analysis of toilet waters, soaps, shower gels, deodorants, sham-
poos, fabric softeners, lipsticks, air freshener, detergents and household cleaning
products, and is also well suited to investigate fragrance profiles from flowers
or essential oils.
One very important advantage of the SPME technique is that it can be
combined with other chromatographic or spectroscopic processes. Thermal de-
sorption of the SPME-concentrated compounds into the carrier gas stream of a
gas chromatograph equipped with an FPD or NPD allows rapid screening for
nitrogen- or sulfur-containing molecules. If SPME is combined with mass spec-
trometry, it is possible to identify any analyte (SPME-GC/MS). If sensory evalua-
tion of the separated flavor compounds is necessary, the SPME method can be
combined with gas chromatographyolfactometry (SPME-GC/O) (Fig. 13). Last,
FIGURE 13 HS-SPME of dry fruit flavor in combination with gas chromatographyolfac-

tometry.
156 Werkhoff et al.
as already mentioned, SPME in combination with enantio-selective GC can sup-

ply valuable information about the enantiomeric composition of chiral flavor
compounds. As an additional example of this, Fig. 14 illustrates enantio-selective
headspace SPME/MDGC analysis of carvone in a toothpaste. The figure shows
that carvone is present in the toothpaste in the (R)-()-formi.e., the enantiomer
that possesses the intense spearmint character.
In conclusion, it should be mentioned that the SPME technique offers a
fast, precise, relatively simple, easily managed, and extremely powerful and in-
formative analytical system for the determination of a broad spectrum of flavor
and fragrance chemicals with varying volatility and polarity properties. Sample
preparation is one of the most critical aspects of the analysis of complex matrices
for trace components and can also be the most time consuming. The introduction
of SPME represented a real advantage in the strategies available to the analyst
faced with problems of sample concentration and cleanup. Another important
step forward in solid-phase microextraction is the fact that the extraction and
desorption steps of SPME-GC have now been fully automated, yielding the addi-
tional capability of series analyses. In such an automated system, the exposure
time of fiber can be precisely controlled. As a result, relative standard deviation
values (% RSD) are less than 5% for most target analytes (71).
FIGURE 14 Headspace SPME/MDGC analysis of a toothpaste to determine the enantio-

meric composition of carvone (PDMS/DVB fiber).
III. HIGH-PRESSURE EXTRACTION WITH

SUPERCRITICAL CO 2
Extraction using supercritical carbon dioxide has been an established industrial-

scale technique for many years. High-pressure CO 2 extraction is already widely
used, for example, for dealcoholization; decaffeination of coffee and tea; pro-
cessing of tobacco, hops, spices, and fats and oils from both vegetable and animal
sources; and also to extract specific compounds or active ingredients for the food,
beverage, and tobacco sector and in the chemical and pharmaceutical industries,
as well as in the fields of cosmetics, leather, textile, paints, and beverages.
In recent years, the elegant and gentle high-pressure extraction method us-
ing supercritical gases has also been introduced in many research laboratories
as a technique for isolating and analytically investigating flavor and fragrance
chemicals from natural products. The resulting flood of scientific publications is
beyond the scope of detailed discussion in this chapter. More detailed information
and theoretical data about extraction with compressed CO 2, may be found in
recent review articles and several monographs (7286).
High-pressure extraction with compressed gases is an effective process for
separating flavor and fragrance chemicals from complex matrices without the use
of organic solvents. The principle of high-pressure extraction is based on the
well-known fact that in the supercritical region, gases can acquire solvent proper-
ties that are superior to those of liquid solvents.
In principle, a variety of gases that can be rendered supercritical in terms
of temperature and pressure could be used as extraction media. In practice, super-
critical CO 2 has proven optimal for the extractive processing of natural sub-
stances. Carbon dioxidea ubiquitous natural materialis physiologically
harmless and inexpensive; is available in high purity; is chemically stable and
inert, nonflammable, and noncorrosive; and has a low critical point. Because of
its low boiling point, CO 2 is easy to remove from both product residues and the
target extracts. An important feature regarding solubility is the polarity of CO 2
under supercritical conditions: CO 2 possesses almost exclusively lipophilic sol-
vent properties, and falls between diethyl ether and methylene chloride in terms
of polarity. It is a highly selective solvent whose dissolving power can be con-
trolled by modifying the temperature and pressure of the extraction process.
When pressure is applied to gaseous CO 2, the result (depending on the
present system temperature) is either liquid or solid carbon dioxide. If a tempera-
ture above the critical temperature is used, however, the gas cannot be liquefied
regardless of how much pressure is applied. The p(T) phase diagram for carbon
dioxide is shown in Fig. 15. Two points on this diagram are of particular interest:
The triple point T p, at which all three phases are present together in equilibrium,
and the critical point K p, which separates the liquid and gaseous phases. This
phase region above the critical temperature and critical pressure of a gasi.e.,
158 Werkhoff et al.
FIGURE 15 p(T) phase diagram for carbon dioxide.
above the critical pointis referred to as supercritical or fluid. Supercriti-

cal or fluid carbon dioxide is considered to exist at pressures above 7.3 MPa and
temperatures above 31.2C. Supercritical gases have excellent dissolving proper-
ties for many classes of chemical substances. The solubility of supercritical CO 2,
for example, is two to three times better than that of subcritical liquid CO 2. The
advantage of compressed gases over organic solvents is that although their den-
sity is very similar to that of liquids, they have a dynamic viscosity that approxi-
mates that of a normal gas. This low viscosity results in rapid mass transport
behavior (high diffusion capability), which allows the supercritical medium to
penetrate particularly easily into the solid material being extracted and dissolve
constituents from it.
With conventional solvents, the solubility of substances can be influenced
primarily by way of temperature; normally, it rises with increasing temperature.
The solubility of organic compounds in supercritical fluids, however, is a function
of temperature and pressure, and can be controlled very easily by modifying the
temperature and pressure. The dissolving power of supercritical fluids generally
rises with increasing pressure (i.e., increasing density). The extraction rate for
natural substances often also depends not only on pressure and temperature but
also on temperature-related diffusion, within the natural matrix, of the constit-
uents being extracted. If the gas density is reduced by modifying the process
parameters (pressure and temperature), the dissolving power decreases so that
dissolved substances are completely or partly separated again. This effect is uti-
lized in the high-pressure extraction separation process. The gas circulates
through the system, which is divided into higher-pressure and lower-pressure
regions. In the loading step (at high gas density), specific substances are dissolved
from the natural matrix. In the demixing step (at lower gas density), the sub-
stances are separated and removed from the circulation loop. High-pressure ex-
traction with supercritical CO 2 avoids elevated temperatures that can lead to
thermal decomposition or rearrangement of labile constituents. Because high-
pressure extraction also avoids any oxidative processes, the method yields high-
quality genuine extracts that are similar to the natural product, retaining almost
the entire native chemical composition of the ingredients. In the field of flavors
and essential oils, extraction with supercritical CO 2 therefore offers extremely
attractive parameters. Under conditions of low thermal stress, the method can
produce concentrates that are usually far superior to conventionally produced
extracts. By adding modifying agents (such as acetone) to supercritical CO 2, it
is possible to vary the dissolving power of the fluid phase over a broad range,
allowing the extraction profile to be modified or extraction yields to be signifi-
cantly enhanced, for example.
The principle and procedure of high-pressure extraction with supercritical
gases is illustrated in Fig. 16. The extractor contains the solid natural material
being processed. Once the entire system has been flooded with CO 2, the pre-
selected supercritical pressure and temperature conditions are established in the
extraction medium using a pump or compressor and a heat exchanger. The gas,
loaded with extracted compounds, is then transported into the separation vessel
where it undergoes changes in pressure and/or temperature, thus demixing and
removing the constituents dissolved in the fluid phase. After separation of the
water that is also extracted (since water becomes increasingly soluble in supercrit-
FIGURE 16 Schematic layout of a high-pressure system for extracting solids with super-
critical gases.
160 Werkhoff et al.
ical CO 2 as the temperature is raised), the method yields flavor extracts with very
good organoleptic properties that can then be analyzed by gas chromatography.
The extract-free gas is drawn off, brought back to the supercritical state, and
returned to the extraction vessel. The process is continuous in terms of the sol-
vent, but discontinuous (batch process) in terms of the solid starting material that
is used, because the introduction of solids into the high-pressure area is problem-
atic.
Liquid products can, of course, also be processed with this extraction tech-
nique; in this case the liquid raw material is fed into a distillation column and
slowly falls to the bottom in countercurrent against the rising supercritical gas.
At the bottom of the column, the residue is discharged from the sump by a level
regulation system, and the extracted components are once again precipitated in
the separator. With this procedure, the extractor in Fig. 16 must be replaced by
a countercurrent extraction column. This process can be run continuously because
liquid can be metered into the high-pressure area without difficulty.
SFE has already been used for a number of years for the extraction of food
constituents. There are numerous articles that deal with the advantages, applica-
tions, and possibilities of SFE in flavor analysis (8794).
An extraction procedure performed on kiwi fruits and leek with supercriti-
cal carbon dioxide in our pilot plant (SITEC-Sieber Engineering AG, CH-Maur/
Zurich) at 9.0 MPa and 40C (gas density approximately 0.5 g/cm3) yielded flavor
concentrates that were judged, in an organoleptic evaluation, to be of much better
quality than extracts produced with conventional methods. The capillary gas chro-
matograms for both total flavor extracts are shown in Figs. 17 and 18. Figure 18
shows that numerous sulfur-containing flavor compounds are particularly sig-
nificant in the leek flavor concentrate, contributing to the typical and characteris-
tic overall flavor impression. Extraction using compressed carbon dioxide makes
it possible to recover fruit and vegetable flavor compounds or spice ingredients
under mild process conditions at good yield and with outstanding quality.
On the other hand, extraction with supercritical CO 2 is also a powerful tool
in essential oil and fragrance research, especially because the isolation of volatile
components from flowers using traditional methods such as steam distillation or
dynamic headspace sampling produce fragrance extracts that do not reflect the
organoleptic properties of the natural material. Due to the mild and gentle extrac-
tion conditions, supercritical fluid extraction is most suitable to isolate essential
oils from spices, flowers, blossoms, herbs, leaves, seeds, and roots (95102).
These extracts are generally considered to be organoleptically superior to concen-
trates produced by traditional techniques. Because most of the essential oils from
such natural products are susceptible to thermal degradation, SFE is especially
advantageous for extracting critical and thermally labile fragrance components.
In our laboratory, for example, extraction of roses have been investigated
as follows: 150 g of rose leaves was filled into the extractor and extracted for 2
FIGURE 17 Gas chromatographic separation of a carbon dioxide kiwi-fruit flavor concen-

trate.
FIGURE 18 FID and FPD capillary GC of a high-pressure CO 2 leek extract.

162 Werkhoff et al.
hours with a CO 2 flow rate of 20 kg/h at 9 Mpa and at a temperature of 40C.

The flavor and fragrance of the carbon dioxide rose extract was very close in
quality to actual rose leaves compared with rose extracts isolated by conventional
techniques. A CO 2 extract was obtained that quite faithfully reflected nature be-
cause hydrolysis, oxidation, or thermal changes are completely avoided.
The FID chromatogram of the supercritical carbon dioxide rose extract is
illustrated in Fig. 19. GC/MS analysis has been used to identify components
responsible for rose fragrance. The major compounds were citronellol, geraniol,
and nerol. More detailed analytical results are given in Fig. 19. Furthermore, it
is worth mentioning that -damascenone could not be detected in the CO 2 sample.
According to Surburg et al., this is due to the extremely mild extraction condi-
tions, thus minimizing the formation of artifacts (103). A certain disadvantage
of the CO 2 extraction, however, is the presence of many compounds that do
not contribute to the rose fragrance impression, such as paraffins or long-chain
paraffinic alcohols.
In conclusion, supercritical CO 2 offers considerable advantages as extrac-
FIGURE 19 Capillary GC-FID pattern of rose petals extracted with supercritical CO 2.

Key to compounds identified: 1, citronellyl acetate; 2, geranyl acetate; 3, citronellol; 4,
nerol; 5, 3,5-dimethoxytoluene; 6, geraniol; 7, nonadecane; 8, (Z)-9-nonadecene; 9, 7,8-
dihydro--ionol; 10, eugenol.
tion solvent for food and plant samples. It is apparent that high-pressure extrac-
tion with compressed gases expands the scope of conventional extraction pro-
cesses. It combines the principles of the two conventional separating methods
(extraction and distillation), but exhibits none of their disadvantages.
High-pressure extractionalso called destractionusing natural car-
bon dioxide thus constitutes a true alternative to conventional extraction tech-
niques using organic solvents. The CO 2 extracts consist entirely of the native
constituents and represent the complete spectrum of volatile compounds present
in the natural product.
IV. SOLVENT-ASSISTED FLAVOR EVAPORATION (SAFE)
Through the years, analytical chemists have made great efforts to develop effi-
cient, reliable methods to extract volatile components of interest from a variety
of complex food matrices. Various extraction methods exist. Each technique has
its particular advantages and disadvantages. The method of choice strongly de-
pends on the variety and complexity of the matrix and the intended analytical
investigations.
Solvent-assisted flavor evaporation (SAFE)a new and versatile technique
for the careful and direct isolation of aroma compounds from complex food matri-
ceswas developed in 1999 by W. Engel et al. (104). Technical details and the
design of the SAFE apparatus are exhaustively described by the authors. SAFE
includes a compact distillation unit in combination with a high vacuum pump
(Fig. 20). The developers of this new equipment pointed out the following advan-
tages:
Higher yields of volatiles compared with the previously used high vacuum
transfer technique
Higher yields of more polar flavor substances
Higher yields of odorants from fat-containing matrices
Direct distillation of aqueous samples such as milk, beer, orange juice, fruit
pulps
Recovery of really authentic flavor extractsi.e., the new extraction proce-
dure provides a flavor sample with organoleptic properties as close as
possible to the natural product
Reliable quantification of polar and labile trace volatiles in complex matri-
ces compared with many other highly sophisticated modern isolation
methods
In this section, the application of SAFE for the analysis of cheese aroma and the
analysis of volatile fragrance components in washing powder is demonstrated.
164 Werkhoff et al.
FIGURE 20 Schematic diagram of the basic components of the SAFE technique.
Furthermore, the direct distillation of grapefruit juice as well as of a soft drink

is described.
A. Extraction of Cheese
The cheese sample was frozen in liquid nitrogen, then broken into smaller pieces
and ground in a Grindomix blender. Powdered material was suspended in diethyl
ether and the suspension was stirred for 18 hours at room temperature. The etheral
solution was dried over anhydrous Na 2SO 4 and concentrated to a volume of 200
ml by means of a Vigreux column. Separation of volatile components from non-
volatile cheese materials was performed by using the SAFE technique. Subse-
quently, the total flavor extract was separated in basic, neutral, and acidic frac-
tions. Nitrogen-containing components were extracted with aqueous HCl and
carboxylic acids were removed by extraction with aqueous sodium carbonate.
Figure 21 shows a comparison of the neutral cheese fractions isolated not
only by SAFE but also by the SDE technique. There were differences in both
the qualitative and quantitative composition. Many heat-induced alterations of
the aroma profile in the SDE sample were observed (data not presented). Identifi-
cation of cheese volatiles was based on GC/MS. We found, based on our spectro-
scopic investigation of the acidic fraction, that the SAFE technique was more
effective in extracting less volatile and polar constituents such as 4-hydroxy-
2,5-dimethyl-3(2H)-furanone, 4-hydroxy-5-methyl-3(2H)-furanone, and 5-ethyl-
4-hydroxy-2-methyl-3(2H)-furanone. As a result, it should be possible to quantify
these organoleptically important trace constituents by special techniquese.g.,
FIGURE 21 Comparison of neutral volatile cheese components isolated by SDE (top) and
SAFE (bottom).
166 Werkhoff et al.
isotope dilution assay using labeled components as internal standards. Further-

more, it is worth mentioning that the SAFE sample revealed the more authentic
and representative aroma of the original product.
B. Direct Distillation of Beverages

Commercially available grapefruit juice and a soft drink were directly distilled
by means of the SAFE technique. In this way, aqueous distillates free from any
nonvolatile material were obtained. After extraction of the aqueous distil-
lates with pentane/diethyl ether 1: 1, flavor concentrates ready for direct HRGC
and GC/MS were produced. The resulting chromatograms are shown in Fig. 22.
We assume that all important volatile compounds were recovered by the aid of
the SAFE distillation and the following extraction step because the aromatic ex-
tracts were truly representative. The soft drink extract was characterized by
typical citrus oil notes, and the grapefruit extract provided characteristic bitter,
fruity, green, tropical, juicy, citrus-like, and clear grapefruit-like aroma impres-
sions.
In Fig. 23, both SAFE extracts are depicted again but have been separated
by means of a relatively new GC technique: EZ-Flash, an assembly that can be
used to upgrade existing gas chromatographs. EZ-Flash is an innovative chro-
matographic system that accomplishes in a few minutes what traditional GC does
in an hour or morei.e., Flash GC is over 20 times faster than conventional GC.
The Flash column consists of a conventional column inserted into a metal tube,
which is resistively heated by a precision power supply. The combination of a
short capillary column, a high gas flow rate, and fast temperature programming
significantly decreases analysis times. The application of EZ-Flash, however,
causes some loss of separation efficiencyi.e. in most cases, resistive heating
cannot be used to reduce the analysis time of extremely complex flavor or fra-
grance samples without causing some loss of separation efficiency. Nevertheless,
in spite of this disadvantage, the approach is ideally suited for rapid screening
as illustrated in Fig. 23. The fundamental principle and some specific applications
of this new technique have recently been published (105108).
As we have already emphasized in this chapter, aroma profiles depend on
the extraction method used. There is no universal extraction procedure in food
flavor analysis, and different methods will influence aroma profiles. The new
SAFE technique presented above represents an attractive approach in flavor re-
search and a new tool for isolating and concentrating volatile components from
solid or aqueous food materials. For that reason, the SAFE technique is expected
to be widely applied in the future for highly efficient extraction of flavor compo-
nents from various food samples.
FIGURE 22 Direct distillation of beverages by means of the SAFE technique (30 m
0.25 mm I.D. DB-WAX; 0.25 m d f; 60C-3C/min 230C). Key to components identi-
fied in soft drink: 1, limonene; 2, -terpinene; 3, -terpinolene; 4, nonanal; 5, linalool;
6, fenchol; 7, 1-terpinen-4-ol; 8, -terpineol; 9, cinnamaldehyde; 10, myristicin. Key to
components identified in grapefruit juice: 1, limonene; 2, cis-linalool oxide (f); 3, trans-
linalool oxide (f); 4, -caryophyllene; 5, -terpineol; 6, trans-carveol; 7, dihydronootka-
tone; 8, nootkatone.
168 Werkhoff et al.
FIGURE 23 10 m EZ Flash GC-FID chromatograms of soft drink and grapefruit juice

extracts (10 m 0.1 mm I.D. DB-WAX; 0.2 m d f; 50C90C/min 230C). Identifi-
cation, see Fig. 22.
TABLE 1 Recovery Data Obtained by Distilling a Solvent Solution of 17 Perfume Oil

Components as well as a Pentane/Diethyl Ether Extract from Washing Powder
Recovery (%)
Model mixture in Model mixture extracted
Compound pentane/diethyl ether from washing powder
cis-3-Hexenyl acetate 73.6 50.5
2,4-Dimethyl-3-cyclohexene-1- 76.6 62.9
carboxaldehyde (Vertocitral)
Linalool 75.6 57.4
-Citronellyl acetate 77.9 54.9
Styrallyl acetate 77.1 77.0
Benzyl acetate 76.4 48.6
-Geranonitril 78.9 65.6
-Citronellol 77.3 69.8
Benzyl acetone 79.9 51.4
Lilial 80.4 46.6
Isocylemone E 78.9 64.9
Sandolene 80.7 66.9
Hexyl salicylate 79.5 76.4
Acetyl cedrene 79.5 59.5
-Hexylcinnamaldehyde 79.8 45.1
Coumarin 79.5 64.6
Methyl -naphthyl ketone 79.6 53.9
C. Extraction of Washing Powder

The analysis of fragrance volatiles in washing powder is a further excellent exam-
ple of the potential of the SAFE technique. In order to test the efficiency of the
SAFE technique with regard to perfume oil components, a model mixture of
fragrance compounds dissolved in pentane/diethyl ether was distilled by means
of the new apparatus. In addition, to study the influence of the washing powder
matrix on the yields of the fragrance volatiles, the model mixture was incorpo-
rated in the washing powder, extracted with pentane/diethyl ether, and distilled
again using the SAFE equipment. As shown in Table 1, no fragrance component
was completely recovered from the organic solution. As expected, investigating
the influence of nonvolatile washing powder materials, the yield of each fragrance
substance was decreased once again compared to the pure pentane/diethyl ether
solution. On the whole, however, data in Table 1 show the SAFE technique to be
a reliable and useful method for isolating fragrance components from perfumery
matrices.
170 Werkhoff et al.
FIGURE 24 Total ion chromatograms of washing powder. (A) SAFE versus (B) SDE.
Key: 1, Prenyl acetate; 2, cis-3-Hexenyl acetate; 3, 2-tert.-Butylcyclohexyl acetate (cis-
Agrumex); 4, 2-tert.-Butylcyclohexyl acetate (trans-Agrumex); 5, 1-Phenylethyl acetate;
6, Geranyl acetate; 7, Verdyl acetate.
1. SAFE Versus SDE

The fragrance of washing powder was analyzed by means of SAFE and compared
with a fragrance extract obtained by SDE. The simultaneous distillation-extrac-
tion (SDE) under atmospheric pressure is a well established and common tech-
nique in essential oil and fragrance research. SDE is an elegant and fast extraction
method and gives good recoveries of many steam-distillable volatiles. However,
polar and very water soluble components are poorly recovered. Heat-induced
artifact formation and decomposition of labile components are further drawbacks
of this technique. The advantage of direct solvent extraction in combination with
the SAFE technique is that it is capable of isolating a wide range of compounds
including very water soluble substances. In addition, heat-induced changes in the
pattern of volatiles due to elevated temperatures applied during SDE are com-
pletely eliminated.
As shown in Fig. 24, the compositions of both extracts clearly differed both
qualitatively and quantitatively. One remarkable difference is the presence and
high concentration of esters in the SAFE sample. Especially all esters are recov-
ered in good yield by the SAFE technique while this class of compounds is clearly
discriminated by the SDE method. Probably some hydrolysis of the esters occurs
during the distillation step at elevated temperature. Obviously, only cis- and
trans-Agrumex are stable during simultaneous distillation-extraction under atmo-
spheric pressure (Fig. 24). Even compounds as nonvolatile as -hexylcinnamal-
dehyde, methyl -naphthyl ketone, coumarin, methyl anthranilate, or methyl di-
hydrojasmonate could be readily detected in the SAFE sample in small amounts.
These odoriferous trace components impart a typical fragrance note to the SAFE
sample and are responsible for the sensorial differences of both distillates to some
extent. A further serious disadvantage of SDE is the fact that many additional
undesirable volatiles are also extracted from the washing powder matrix in high
concentration (e.g., tridecanol, tetradecanol, pentadecanol, branched alcohols,
alkoxy alcohols), overlapping with fragrance components and thus, of course,
complicating the interpretation of analytical results.
In summary, the SAFE technique has been shown to extract volatile compo-
nents efficiently from perfumery matrices. This new technique uses moderate
extraction temperatures that do not degrade labile fragrance molecules or produce
artifacts due to the analytical technique. Therefore, this method can advanta-
geously be used in fragrance analysis in place of classic sample preparation tech-
niques in future.
V. STIR BAR SORPTIVE EXTRACTION (SBSE)
Recently, a novel technique for isolation and concentration from aqueous sam-
ples, namely Stir Bar Sorptive Extraction (SBSE), was developed by E. Baltussen
et al. (109). SBSE has already proven to be a very versatile technique for analyti-
cal chemistry. To illustrate the potential of SBSE for volatile analytes, this new
technique in combination with thermal desorption and capillary gas chromatogra-
phy coupled to a mass spectrometer was used for enrichment of lemon-flavored
beverages (110).
The stir bar (Gerstel Twister) consists of a magnetic core sealed inside
a glass tube and is coated with a thick layer (0.5 mm or 1 mm) of polydimethyl-
siloxane (PDMS). The extraction mechanism is very similar to solid-phase mi-
croextraction (SPME), i.e., concentration occurs by sorption of the analytes in
the polymeric liquid PDMS phase. The stir bar is introduced in the aqueous sam-
ple, and extraction takes place during stirring. Subsequently, the stir bar is re-
moved from the aqueous sample, introduced in an empty thermal desorption glass
tube, and transferred to a thermal desorption unit where the analytes are thermally
released and transferred to the GC system. A schematic diagram of the analytical
system is shown in Fig. 25. The instrumental set-up consists of a Gerstel TDS
2 thermal desorption system mounted on a HP 6890 gas chromatograph coupled
172 Werkhoff et al.
FIGURE 25 A schematic view of the SBSE concentration device.
to a mass selective detector. Conditioning of the PDMS coated stir bar, extraction
procedure, and thermal desorption are described in detail by Tredoux et al. (110).
Figure 26 shows the chromatograms obtained from the analysis of 5 ml of
grapefruit juice and 5 ml of whisky. Many nonpolar and semipolar flavor sub-
stances are enriched by the SBSE technique. Figure 26 clearly illustrates the
excellent performance for volatile and semivolatile flavor components, which
appear as sharp and symmetrical peaks. It was found that nearly no interferences
were detected when the stir bar was directly exposed to the cloudy juice matrix.
In Fig. 27 the chromatographic comparison of a soft drink sample obtained
by SBSE and SPME extraction is shown. Although the gas chromatographic con-
ditions were completely different for both sampling techniquesand therefore
direct comparison is not possiblestriking differences in the recoveries of flavor
components can be observed.
In addition, we have modified the original SBSE concept and have applied
the Gerstel Twister to headspace sampling for the first time. In Fig. 28, a
comparison of the chromatograms obtained by SBSE from both the aqueous and
gaseous phase of a soft drink sample is depicted. Fascinating differences in the
FIGURE 26 Stir Bar Sorptive Extraction (SBSE) applied to grapefruit juice (top) and
whisky (bottom). TDS: temperature: 20C/25C/min to 250C/5 min; transfer line:
300C; purge flow: 25 ml/min hydrogen; TDS-mode: splitless. CIS: split-mode: 1 :5; tem-
perature: 120C/12C/sec to 300C; carrier gas: 1.3 ml/min hydrogen; column: 20 m
DB-WAX; 0.18 mm I.D.; film thickness: 0.18 m; temp.-progr.: 40C-1 min isoth.-9C/
min to 250C.
recoveries of higher boiling or more polar flavor substances can be noticed (top
and middle chromatogram). Due to the analytical differences obtained in Fig. 28,
we felt encouraged to combine both techniques in order to extract the aqueous
as well as the gaseous phase at the same time, working with two stir bar systems.
Subsequently, both stir bars were introduced into the empty glass tube of our
desorption system and simultaneously thermally desorbed. Making use of this
procedure, comprehensive information on the flavor composition of a beverage
may be obtained as depicted in Fig. 28 (bottom chromatogram: SBSE2). SBSE2
includes both headspace SBSE and immersion SBSE, and as a result an additive
chromatographic pattern is obtained. From the bottom chromatogram in Fig. 28, it
is immediately clear that a certain range of analytes including cinnamic aldehyde,
cinnamyl acetate, -bisabolol, myristicin, campherenol and 2-methoxycinnamic
174 Werkhoff et al.
aldehyde is significantly better recovered from the aqueous sample while higher-
boiling substances with long-chain hydrophobic groups could preferably be re-
covered by headspace SBSE (e.g., myristic acid, pentadecanoic acid, palmitic
acid, squalene). Consequently, SBSE2 is a useful, practical, versatile, and univer-
sal technique and may be successfully used to analyze flavor and fragrance com-
pounds from aqueous samples with high selectivity and specificity in future.
Figure 29 represents the enrichment of volatile flavor components from
one whole grape. To this end the stir bar was inserted directly into the fruit pulp
of the grape. The absorbed flavor compounds were thermally desorbed and were
transferred directly to the column for analysis.
Figure 30 shows the application of HS-SBSE to cotton towels that have
been washed in the presence of a commercially available liquid washing agent
and then dried in open air. The aim of this experiment was to study the adherence
of perfume oil components at the surface of textiles. In advance, it should be
mentioned that this fragrance analysis was not possible with HS-SPME due to
lack of sensitivity. Because of the much higher PDMS film thickness of the stir
bar, however, the sensitivity using HS-SBSE could be drastically increased. Some
pieces of a cotton towel were placed in a round-bottomed glass flask together
with the PDMS coated stir bar. Sampling was performed at room temperature
over a 18-hour period. Subsequently, the stir bar was transferred into the empty
glass thermal desorption tube and desorbed into a gas chromatograph coupled
with mass spectrometry. The experimental conditions were the same as those
described in Fig. 26.
A series of saturated aldehydes from C 6-C 10 were detected. This is not un-
common because aldehydes are usual constituents of perfume oil compositions.
A further explanation for all the aldehydes on the cotton towel, however, could
be found in a paper of B.D. Mookherjee and R.W. Trenkle, who undoubtedly
demonstrated that cotton in the presence of air and sunlight generates this homol-
ogous series of aldehydes (111). In addition, numerous perfume oil components
could unequivocally be identified that obviously possess good adhesive power
on the surface of the analyzed textile. Main components are numbered in Fig.
30. In total, identification of more than 40 volatile perfume oil components was
possible (data not presented here).
In conclusion, the present studies demonstrate the practicability of SBSE
FIGURE 27 Comparison of volatiles extracted from a soft drink with SPME (upper curve)
and SBSE (lower curve); the SPME fiber used was PDMS/DVB. (A) 5 ml soft drink
sample. Immersion SPME at room temperature using a 65 m PDMS/DVB fiber, column:
30 m DB-WAX, 0.25 mm I.D. 0.25 m film thickness. (B) 5 ml soft drink sample. SBSE
at room temperature. Column: 20 m DB-WAX, 0.18 mm I.D. 0.18 m film thickness.
FIGURE 28 Comparison of chromatograms of soft drink extracted from both the aqueous
(top) and gaseous phase (middle). Combined sampling technique: SBSE2 (bottom). Peak
identities are as follows: 1, terpinolene; 2, nonanal; 3, linalool; 4, 1-terpinen-4-ol; 5, -
terpineol; 6, geranyl acetate; 7, safrol; 8, cinnamic aldehyde; 9, cinnamyl acetate; 10, -
bisabolol; 11, myristicin; 12, campherenol; 13, 2-methoxycinnamic aldehyde; 14, myristic
acid; 15, pentadecanoic acid; 16, palmitic acid; 17, squalene; 18, caffeine.
Isolating and Quantifying Volatile Compounds
FIGURE 29 Direct SBSE sampling of volatile flavor components from one whole grape. Peaks are as follows: 1, hexanol; 2, 2-nonanone;
3, furfural; 4,5-methylfurfural; 5, phenylacetaldehyde; 6, -hexalactone; 7, cyclotene; 8, -octalactone, 9, 2-phenylethyl alcohol; 10, furaneol;
177
11, p- and m-cresol; 12, norfuraneol; 13, -decalactone; 14, 3,5-dihydroxy-6-methyl-2,3-dihydro-pyran-4-one; 15, 2-formyl-5-hydroxymeth-
ylfuran.
178 Werkhoff et al.
FIGURE 30 Adherence of odorous substances from a washing agent to textile surfaces.

Headspace SBSE/HRGC/MS chromatogram of volatile perfume oil components extracted
from a cotton towel with the PDMS stir bar. Peak identities are as follows: 1, hexanal;
2, heptanal; 3, octanal; 4, nonanal; 5, (E)-2-octenal; 6, tetrahydrolinalool; 7, decanal; 8,
(E)-2-nonenal; 9, (E)-2-decenal; 10, Oryclon; 11, -terpineol; 12, verdyl acetate P1; 13,
verdyl acetate P2; 14, iso-E-Super; 15, pentyl salicylate; 16, hexyl salicylate; 17, acetyl
cedrene; 18, Chromanolide P1; 19, Chromanolide P2; 20, -hexylcinnamic aldehyde.
in the qualitative analysis of soft drinks, fruit juices, and alcoholic beverages.
Obviously, the potential of SBSE for determining individual organic compounds
in aqueous matrices is good. SBSE seems to be a reliable, versatile, and extremely
sensitive technique for the analysis of flavor compounds in aqueous or other
liquid matrices.
Here again, it is important to emphasize that SBSE is extremely sensitive
due to two facts. On the one side, it is well-known that thermal desorption ensures
a very high sensitivity. On the other side, however, high concentration factors
by the thick PDMS layer of the stir bar are achieved. Due to the substantial
increased amount of PDMS in SBSE (compared with SPME) a significant in-
crease in sensitivity can be additionally attained. For this reason, SBSE shows
tremendous promise as a versatile tool with especially high sampling sensitivity
and should be applicable as an enrichment system for a wide variety of sample

types and analytes.
Moreover, quantitative analysis of volatile organic substances in aqueous
matrices using isotopically labeled internal standards should also be easily
achieved with high accuracy. Therefore, SBSE is expected to be the first-choice
sampling technique for many flavor and fragrance applications in different matri-
ces in future. As SBSE is still in its infancy, we are anxious to see if stir bars
coated with medium polar or polar sorbents will be developed in the near future
and if this new and powerful sample clean-up procedure will be extended to
entirely other fields of application.
VI. STABLE ISOTOPE DILUTION ASSAYA HIGHLY

PROMISING METHOD FOR QUANTIFYING FLAVOR
AND FRAGRANCE COMPOUNDS
The objective of modern flavor analysis is to qualitatively and quantitatively de-

termine the flavor profile imparted by nature and thus to identify those substances
responsible for the flavor of a food. In recent years, a combination of sensory
methods and instrumental analytical procedures has made a major contribution
toward reaching this objective. Thanks in particular to the concept of odor activity
value, which has been steadily refined at the German Research Institute for Food
Chemistry in Garching (Technical University of Munich) during the past decade,
it is today possible to make well-founded statements about the odor activity of
volatile flavor compounds in foods (112116). With the aid of the odor activity
value concept, it is possible to identify those components of the extremely com-
plicated flavor profiles that are of key importance for the specific flavor of a food.
Odor activity value (OAV) is defined as the ratio of concentration to odor
threshold. Hence, to be able to calculate odor activity values, it is first necessary
to exactly quantify the compounds that are deemed to be of interest in a food.
It is relatively difficult to provide an exact quantitative analysis of com-
pounds that are unstable and occur only in traces in foods. Serious errors are
possible, particularly when the volatile flavor compounds are present in extremely
small amounts, are unstable, and/or are reactive, and the workup conditions se-
lected often result in considerable losses. It is known that quantitative results
obtained for flavor compounds in foods can be significantly affected by the isola-
tion method and the structure of the analyte. Quantitative analysis can readily
give erroneous results, particularly when the internal standards that are used differ
from the analytes in terms of their chemical and physical properties, as is usually
the case. Although the method of quantifying volatile compounds with the aid
of internal standards has been in use for quite some time, the literature repeatedly
180 Werkhoff et al.
discloses critical opinions drawing attention to the inadequacies and sources of

error of this method (117). The use of internal standards in quantitative flavor
analysis makes it absolutely essential to determine the percentage recovery and
GC response factors, which can involve considerable analytical errors (118). To
minimize these errors, different research groups in the past often simultaneously
used several internal standards having different functional groups for quantita-
tively determining flavor compounds in foods (119123).
The method of choice for exact quantitative determination of key flavor-
relevant constituents in foods is currently IDA, isotope dilution analysis (stable
isotope dilution assay or SIDA), in which the isotope-labeled analog used as the
internal standard undergoes the same losses during workup as does the analyte.
Either radioactive or stable isotopes can be used for the isotope dilution method.
The latter are simpler and can be handled without health risks. For this reason,
only stable isotopes are used in flavor analysis.
Isotopes are nuclides that have the same number of protons but a different
number of neutrons in the nucleusi.e., they have different mass numbers. The
atomic nucleus of the 12C carbon isotope consists of 6 protons and 6 neutrons,
whereas that of the 13C isotope consists of 6 protons and 7 neutrons. The atomic
nucleus of the 1H hydrogen isotope consists of one proton, whereas the 2H hydro-
gen isotope (deuterium) contains an additional neutron. Because the isotopes of
a chemical element have the same number of protons in the nucleus, they also
have the same number of electrons in the shell. Hence, they display the same
chemical behavior.
The principle of isotope dilution analysis consists of synthesizing the flavor
compounds to be quantified (native compounds) so as to label them with stable
isotopes (for example, deuterium or carbon-13) and adding them to the food
matrix before sample workup.
The flavor compounds that are labeled with stable isotopes (isotopomers)
differ only slightly from the analyte in terms of mass, and their physical and
chemical propertiese.g., volatility, reactivity, distribution coefficient and chro-
matographic behaviorare the same as those of the unlabeled flavor compounds,
with the exception of minor and negligible isotope effects. They are added to
foods as internal standards as early as possible, namely before the first extraction,
so that they undergo virtually the same losses as the flavor compounds to be
studied during the isolation method and enrichment steps that are employed. For
this reason, labeled compounds satisfy nearly all of the requirements for an ideal
internal standard and can also tolerate workup methods with very low recovery
percentages, provided that the detection sensitivity is not too low.
A fundamental prerequisite for absolutely correct quantitative values is
thoroughly homogeneous distribution of the labeled standard and analyte in the
sample. In this regard, slight differences in adsorption characteristics between
the analyte and the labeled standard must be tolerated, because the native com-
pounds that are present in the sample matrix are bound to surfaces in a manner
that differs from those in the added isotope-labeled standard. With the aid of the
known amount of added labeled standard and the MS response factorwhich
is determined at defined weight ratios of analyte to standard under identical GC/
MS conditionsit is possible to exactly calculate the concentration of flavor
compounds in foods.
Isotope labeling means the deliberate incorporation of specific isotopes into
an organic molecule. Deuterium and carbon-13 are the isotopes that are most
frequently used in flavor analysis. Deuterium atoms, however, must not be intro-
duced at positions of the molecule at which a D/H exchange can take place.
For example, protons in the -position to a carbonyl group can enolize during
quantification and then undergo an exchange with protons from the sample, pro-
ducing erroneous results. In all analytical work, it is necessary to check for possi-
ble D/H exchange in a labeled molecule by performing exchange tests on model
mixtures under identical workup conditions; otherwise, the results of the quantita-
tive analysis can be significantly falsified.
Differentiation between labeled and unlabeled flavor compounds is based
on the difference in nuclear masses of the isotopes and on employment of the
mass spectrometry method. One of the main advantages of isotope dilution analy-
sis is the high selectivity of the gas chromatography/mass spectrometry combina-
tion. The selectivity and the sensitivity of the determination can be further en-
hanced by the use of different ionization techniques. In most cases, differentiation
between the internal standard and the analyte is achieved with the aid of chemical
ionization with reactant gases, with the intensities of the molecular ions typically
being measured by selective ion monitoring (SIM mode).
A major advantage of stable isotope-labeled flavor compounds is that their
mass spectra are closely comparable with the spectra of the analytes and are
shifted only in accordance with the particular number of deuterium or C-13 atoms
incorporated into the molecule. The labeling should increase the molecular
weight of the internal standard by at least two mass units to minimize influencing
or overlapping with the masses of the analyte. The incorporation of at least three
isotope atoms would be optimal to truly ensure that the ions to be evaluated are
sufficiently separated from the natural isotope ions of the analyte (124). The GC
retention times of the labeled standard and of the analyte are usually very close
(in most cases, the two compounds are even co-eluted during gas chromatogra-
phy), so that no changes are to be expected in the physical parameters of the MS
source during this short time that could affect the evaluation of the ion intensities
and thus the accuracy of the quantitative analysis.
The chemical and isotopic purity of the synthesized labeled standard are
important data and must be known. Ideally, high isotope enrichment is desirable.
Standards of a lower isotopic purity, however, can also be used for quantification
purposes. In this case, evaluation of the data is more complicated, and overlap-
182 Werkhoff et al.
ping of standard and analyte ions must definitely be taken into consideration in
the mathematical calculations.
The principle of isotope dilution analysis is quite old and dates back to a
quantitative analytical method developed by O. Hahn and G. von Hevesy et al.
This method involved adding to the component to be analyzed an exactly defined
amount of the same compound that had been labeled with a radioisotope (radio-
activity measurements of lead and uranium isotopes) (125,126). In 1939, isotope
dilution analysis was introduced into biochemical analysis by R. Schoenheimer
et al. (127) to determine leucine in protein extracts using a reference substance
labeled with a stable isotope as the internal standard. Shortly thereafter, H.G.
Ussing employed this method to determine the leucine content of hemoglobin
(128). One year later, D. Rittenberg and G. L. Foster used the isotope dilution
analysis method for quantitative determination of palmitic acid in rat fat and of
glycine, glutamic acid, and aspartic acid in protein hydrolyzates (129). In 1966,
Ch. Sweeley et al. used the principle of isotope dilution analysis in the quantita-
tive determination of glucose with heptadeuteroglucose as the internal standard
(130). During the following years, the main fields of application for isotope dilu-
tion analysis (IDA) were clinical chemistry and biomedical research (131,132),
veterinary residue analysis (133), the determination of plasticizers in foods
(134,135) and of trace elements in biological materials (136), as well as studies
of patulin and pantothenic acid in foods (137,138) and cork taint in wine (139).
The theoretical basis for mass spectrometric isotope dilution analysis and the
mathematical considerations concerning the correct evaluation of ion intensities
can be found in several articles in the literature (140146).
In the field of flavor research, isotope dilution analysis was first used by
R. L. N. Harris et al. to quantify 2-methoxy-3- isobutylpyrazine in New Zealand
wine (147) and by P. Schieberle and W. Grosch for quantitatively determining
heterocyclic compounds in bread flavor (148). Since that time, the latter research
group has used the odor activity value concept to analyze numerous important
foods whose flavor develops during cooking, baking, frying, or roasting, as well
as fruits, herbs, vegetables, and beverages. Surprisingly, the results, based on
quantitative data and threshold values, showed unequivocally that the number of
flavor-producing compounds in foods is relatively small. The rather comprehen-
sive literature concerning this subject cannot be covered within the scope of this
chapter. Interested readers can find numerous references to the literature on the
Internet at http://dfa.leb.chemie.tu-muenchen.de/DPublikationen.html. For the
sake of completeness, however, it should be noted that the Garching research
group has since developed isotope dilution analysis for more than 100 odor-active
flavor compounds.
In recent years, other research groups have applied the elegant IDA method
to the quantification of volatile compounds in foods (149159). The scope and
the limitations of isotope dilution analysis are discussed in detail in several review
FIGURE 31 Chemical structures of the labeled internal standards used for the quantifica-
tion of volatile flavor compounds. Symbol indicates the labeling position for deuterium.
articles (160162). The main reason that this highly promising and exact method
of quantitative determination of flavor compounds has nevertheless not yet come
into widespread use in analytical laboratories is, in part, the relatively difficult
and, above all, time-consuming and costly synthesis of the isotope-labeled stan-
dards (163172).
We will now use cheese, coffee, apple, a fabric softener, and soap by way
of example to show that the use of isotope dilution analysis permits exact, trouble-
free quantification of fragrance and flavor compounds in the ppm and ppb range.
To this end, the deuterium-labeled internal standards (isotopomers) shown in Fig.
31 were synthesized. Given the scope of this chapter, we have intentionally omit-
ted a discussion of the synthesis of labeled flavor compounds.
The flavor extracts were analyzed with the aid of multidimensional gas
chromatography in combination with a mass spectrometer. We used the MCS 2
184 Werkhoff et al.
FIGURE 32 Calibration curve for the quantitative determination of methional.
column switching system from the Gerstel Company in combination with an HP

5973 mass-selective detector. El mass spectra were recorded at 70 eV, and PCl
spectra with isobutane as the reactant gas. All quantitative measurements were
performed in the SIM mode.
Calibration curves were constructed, and the MS response factors were
determined for all seven flavor compounds. Figure 32 shows a calibration curve
using labeled and unlabeled methional by way of example. The corresponding
El and Cl spectra are shown in Fig. 33. The calibration curve was obtained with
nine defined mixtures of analyte and labeled standard in a concentration range
from 5 to 95 ng. In the graphic representation, the ratio of the I D m/z 108/I H
m/z 105 ion intensities is plotted against the weight ratio W D /W H. The linearity
is satisfactory; the response factor of methional for a 1 :1 mixture is 0.94.
The response factor is responsible for slight differences in the fragmenta-
tion of the analyte and the standard. This factor depends on the synthesis route
and, in particular, upon the isotopic purity and the number and position of the
deuterium atoms. The correction factor for the internal standard is defined by the
following equation:
methional (ng) I108

Rf
[ H3]-methional (ng) I105
2
Table 2 shows the ions of the analyte and of the standard selected for the
FIGURE 33 El and PCl mass spectra of unlabeled and deuterium-labeled methional.
quantitative determination of the flavor compounds, along with chemical purity,

isotopic purity, retention times, and MS response factors.
Moreover, all seven flavor compounds in each matrix were checked for
deuterium exchange with the sample matrix under the isolation conditions used.
None of the compounds showed any change in the ratio of the relevant ion intensi-
ties, meaning that no exchange reactions had taken place in the course of the
analysis.
A. Cheese Flavor
Analytical studies of Parmesan flavor have shown that methional, ethyl caproate,
and p-cresol give high dilution factors in the isotope dilution analysis of the flavor
extract and, hence, represent important constituents of this type of cheese.
For quantitative determination of these flavor compounds, 400 g of grated
Parmesan cheese was quick-frozen with liquid nitrogen and then ground to a fine
powder. The powdered cheese was extracted overnight with dichloromethane
containing 50 g of the internal standard [2H 3]-methional, 400 g of [2H 5]-ethyl
caproate, and 4 g of [2H 7]-p-cresol. The volatile compounds were distilled from
186
TABLE 2 Chemical and Isotopic Purities, Ion Masses, Retention Times, and MS Response Factors of Analytes and Deuterium-Labeled
Internal Standards
Retention
Chemical purity Isotopic purity Ion mass time min MS response
Analyte ISTD GC-FID % GC/MS-PCl % m/z 30 m DB-WAX Factor R f
Methional 85.0 105 11.37 0.94

d 3-Methional 99.0 99.9[2H 3] 108 11.31
Ethyl caproate 99.4 145 5.51 1.15
d 5-Ethyl caproate 99.0 99.5[2H 5] & [2H 6] 150 & 151 5.44 & 5.41
p-Cresol 99.0 109 32.62 0.82
d 7-p-Cresol 99.0 99.5[2H 7] 116 32.50
Guaiacol 99.5 125 26.70 0.85
d 3-Guaiacol 97.2 99.6[2H 3] 128 26.64
Skatole 99.5 132 45.29 1.08
d 3-Skatole 96.0 99.8[2H 3] 135 45.15
2-Ethyl methylbutyrate 99.7 131 4.23 0.90
d 5-2-Ethyl methylbutyrate 94.0 90.6[2H 5] 136 4.18
2-Phenylethyl alcohol 99.7 105 27.22 1.07
d 2-2-Phenylethyl alcohol 98.0 99.8[2H 2] 107 27.10
Werkhoff et al.
FIGURE 34 El and PCl mass spectra of unlabeled and deuterium-labeled p-cresol.
the nonvolatile cheese constituents by high-vacuum distillation (SAFE apparatus)

at a water bath temperature of 70C (104). To separate the carboxylic acids, the
organic phase was extracted three times with 100-ml portions of 0.5 molar sodium
carbonate solution and then washed neutral with 100 ml of saturated sodium
chloride solution. The flavor solution was dried over sodium sulfate, concentrated
with the aid of a Vigreux column to a final volume of 1 ml, and analyzed by
GC/MS. The El and Cl mass spectra of the three flavor compounds are shown
in Figs. 33, 34, and 35. The mass traces were evaluated at m/z 105 for unlabeled
methional and at m/z 108 for the deuterated compound. The mass traces were
recorded at m/z 145 and m/z 150151 for ethyl caproate, and at m/z 109 and
m/z 116 for p-cresol. The quantitative analysis of Parmesan cheese gave 125
g/kg for methional, 660 g/kg for ethyl caproate, and 12.3 g/kg for p-cresol.
B. Coffee Flavor
More than 800 volatile flavor compounds have thus far been identified in roasted
coffee and in coffee beverages made from it. Individual compounds in this ex-
188 Werkhoff et al.
FIGURE 35 El and PCl mass spectra of unlabeled and deuterium-labeled ethyl caproate.
tremely complicated mixture can be identified exactly and relatively easily with
the aid of IDA. We selected guaiacol and skatole for quantitative determination
in a coffee beverage. To prepare the beverage, 70 ml of hot water was filtered
through 10 g of ground coffee. The SPME technique was used to isolate the
flavor compounds. The labeled internal standards (50 g of [2H 3]-guaiacol and
3 g of [2H 3]-skatole) were added to the coffee infusion (50 ml), which was
contained in a septum-sealed container (100 ml) and adjusted to a temperature
of 60C. The system was equilibrated by stirring for 30 minutes. Sampling from
the systems headspace also lasted 30 minutes and was done with an SPME fiber
having a 65-m thick polydimethylsiloxane/divinylbenzene (PDMS/DVB) coat-
ing. The fiber was then subjected to thermal desorption at 230C in the injector
of a GC/MS system. Figure 36 shows a detail of the GC/MS analysis. The MS
data for guaiacoland the evaluated mass tracesare presented in Fig. 37. The
calculated amount of guaiacol in the coffee beverage was 920 g/L and that of
skatole 64 g/L (see mass spectra in Fig. 40). Recently, SIDA combined with
SPME was applied for the rapid and direct quantification of selected coffee im-
pact odorants (173).
FIGURE 36 Extraction of coffee brew flavor components by means of headspace SPME.

Quantification of guaiacol and skatole.
C. Apple Flavor
The efficacy of IDA will be demonstrated using the example of ethyl-2-methylbu-
tyrate, a trace component of apple flavorone which because of its low sensory
threshold value, however, can persistently affect the flavor.
To isolate the flavor compounds, we used simultaneous distillation/extrac-
tion by the Likens-Nickerson method, because ethyl-2-methylbutyrate is highly
steam-volatile. To 750 g of comminuted apple we added 1.5 L of water containing
7.5 g of [2H 5]-ethyl-2-methylbutyrate as the internal standard. Extractant: Di-
ethyl ether/pentane 1 :1, extraction time: 2.5 hours. The final volume of the flavor
concentrate for GC/MS analysis was 1 ml. Figure 38 shows the El and Cl mass
spectra of ethyl-2-methylbutyrate. Mass traces m/z 131 and m/z 136 were used
for quantitative evaluation in the SIM mode. The analyzed apple contained 3.5
g/kg of ethyl-2-methylbutyrate.
D. Fragrance Analysis in a Fabric Softener and in Soap

We will use 2-phenylethyl alcohol, a component of the perfume oil contained in
a fabric softener, by way of example to demonstrate that IDA is a suitable method
190 Werkhoff et al.
FIGURE 37 El and PCl mass spectra of guaiacol as well as mass traces of the analyte
and labeled internal standard (m/z 125 and m/z 128).
for quantification of fragrance compounds in products from the perfume sector

and cosmetics industryassuming that isotope-labeled fragrance compounds are
available as internal standards.
100 ml of distilled water was added to 50 g of fabric softener, and the
mixture was extracted for 2 hours, with agitation, with 250 ml of dichloromethane
containing 3.75 mg of [2H 2]-2-phenylethyl alcohol. To separate the volatile from
the nonvolatile constituents of the fabric softener, the dichloromethane extract
was fed dropwise into a SAFE apparatus and distilled at 70C. The organic phase
was dried over sodium sulfate, and a final volume of 2 ml was obtained by use
of a Vigreux column. The fragrance compound concentrate was analyzed by GC/
MS. Figure 39 shows the El and Cl mass spectra of 2-phenylethyl alcohol. After
evaluation of the mass traces at m/z 105 and m/z 107, we obtained a value of
74 mg/kg of 2-phenylethyl alcohol. It takes at least one day to quantify 2-phenyl-
ethyl alcohol in a fabric softener using this extraction procedure.
A substantially faster alternative is solid-phase microextraction (SPME).
In recent years, the SPME technique has been used in our laboratory as an effec-
tive method for analyzing fragrance compounds in cosmetic products, fragrant
splashes, detergents, and cleansers. It is an innovative, fully solvent-free sample
preparation method and represents an excellent alternative to the conventional
FIGURE 38 El and PCl mass spectra of unlabeled and deuterium-labeled ethyl-2-methyl-

butyrate.
sample extraction techniques thus far used in the analysis of fragrance compounds
in consumer products. These methods are very time-consuming and as a rule
require larger amounts of samples and solvents. Aside from the time factor, the
main advantage of SPME analysis of fabric softeners is primarily the extremely
small sample of fabric softener needed.
Experimental data
Amount of sample 1 g of fabric softener

Internal standard 75 g of [2H 2]-2-phenylethyl alcohol
Equilibration time 30 min
Sampling from headspace 30 min
Temperature 60C
SPME fiber PDMS/DVB, 65 m
Desorption temperature 230C
192 Werkhoff et al.
FIGURE 39 El and PCl mass spectra of unlabeled and deuterium-labeled 2-phenylethyl

alcohol.
After evaluation of the mass traces at m/z 105 and m/z 107, the amount of 2-
phenylethyl alcohol found in the fabric softener was 76 mg/kg.
There was good correlation between the results obtained with the two ex-
traction methods, but it should be pointed out that quantification of 2-phenylethyl
alcohol in the fabric softener matrix by means of the SPME method was com-
pleted after only 2 hours.
A further example for application of SIDA in fragrance analysis is the
quantification of skatole in soap. For this analysis, the SPME technique using 65
m polydimethylsiloxane/divinylbenzene (PDMS/DVB) coated fiber was ap-
plied. Headspace sample was prepared as follows: 1 g of soap was dispersed in
10 ml of distilled water, spiked with 1 ppm d 3-skatole, and placed in a sealed 100-
ml vial containing a Teflon magnetic stirring bar. The sample was equilibrated at
60C for a period of 30 min. The fiber was carefully introduced in the headspace,
and perfume oil constituents were extracted for 30 min under magnetic stirring.
FIGURE 40 El and PCl mass spectra of skatole along with mass traces of analyte and
deuterium-labeled internal standard (m/z 132 and m/z 135).
The SPME fiber was subjected to thermal desorption in the GC injector system
at a temperature of 230C. Compound identification was based on comparison
of mass spectra with those of authentic reference compounds. Figure 40 shows
the El and PCl mass spectra of unlabeled and labeled skatole along with the
evaluated mass traces of the analyte and labeled internal standard at m/z 132 and
m/z 135. The analyzed soap contained 950 ppb of skatole. The method used
required an analysis time of approximately approximately 2 h.
Here again, it should be emphasized that solid-phase microextraction is a
sensitive tool for qualitative analysis of many different compounds. Quantitative
analysis, however, using external or internal standard calibration is relatively
difficult, expensive, and time-consuming. SPME in combination with isotopically
labeled compounds, however, is an effective, reliable and easy-to-handle tool for
quantitative flavor and fragrance analysis in different matrices. The time saving
and the enormous improvement of quantitative determination of trace compo-
nents through SPME/SIDA could be of significant benefit in the field of flavor
and fragrance research and will provide our industry with new possibilities.
194 Werkhoff et al.
In summary, we can state that isotope dilution analysis permits exact quan-
titative determination of fragrance and flavor compounds, even in matrices of
complicated composition. The quantitative data for key flavor compounds ob-
tained by flavor compound analysis including SIDA enable their odor activity
values (OAV) to be calculated. This provides a reliable basis for producing flavor
compositions that are very similar to their natural counterparts. Due to its accu-
racy and precision, it might be expected that SIDA will become more widely
used and progressively more applied in flavor and fragrance science in future.
ACKNOWLEDGMENTS
The authors are grateful to Dr. Stefan Lambrecht and Dr. Jan Foerstner for the
synthesis of deuterium-labeled flavor components. Furthermore, the authors wish
to express appreciation to Linda Weber, Margit Liebig, and Michael Roloff for
skillful technical assistance and to Ingrid Wengler for patience and diligence in
preparing this manuscript.
REFERENCES
1. H. Sugisawa, Sample preparation: isolation and concentration, Flavor Research.

Recent Advances (R. Teranishi, R. A. Flath, and H. Sugisawa, eds.), Marcel Dek-
ker, Inc., New York, 1981, p. 11.
2. H. Sugisawa, Prerequisites for sample preparation in flavour analysis, Analysis of
Volatiles (P. Schreier, ed.), Walter de Gruyter, Berlin, New York, 1984, p. 3.
3. M. M. Leahy and G. A. Reineccius, Comparison of methods for the isolation of
volatile compounds from aqueous model systems, Analysis of Volatiles (P.
Schreier, ed.), Walter de Gruyter, Berlin, New York, 1984, p. 19.
4. P. Schreier, Isolierung von Aromastoffen. Entscheidender Schritt bei der Analyse
von Lebensmittelaromen, Lebensmittelchem. Gerichtl. Chem. 41:25 (1987).
5. R. Teranishi and S. Kint, Sample preparation, Flavor Science. Sensible Principles
and Techniques (T. E. Acree and R. Teranishi, eds.), American Chemical Society,
Washington, DC, 1993, p. 137.
6. G. Reineccius, Biases in analytical flavor profiles introduced by isolation method,
Flavor Measurement (C. T. Ho and Ch. H. Manley, eds.), Marcel Dekker, Inc.,
New York, 1993, p. 61.
7. R. G. Buttery and L. C. Ling, Methods for isolating food and plant volatiles, Bio-
technology for Improved Foods and Flavors (G. R. Takeoka, R. Teranishi, P. J.
Williams, and A. Kobayashi, eds.), American Chemical Society, Washington, DC,
1996, p. 240.
8. A. Krumbein and D. Ulrich, Comparison of three sample preparation techniques
for the determination of fresh tomato aroma volatiles, Flavour Science. Recent De-
velopments (A. J. Taylor and D. S. Mottram, eds.), Hartnolls Ltd., Bodmin, UK,
1996, p. 289.
9. Th. H. Parliment, Solvent extraction and distillation techniques, Techniques for Ana-
lyzing Food Aroma (R. Marsili, ed.), Marcel Dekker, Inc., New York, 1997, p. 1.
10. Th. P. Wampler, Analysis of food volatiles using headspace-gas chromatographic
techniques, Techniques for Analyzing Food Aroma (R. Marsili, ed.), Marcel Dek-
ker, Inc., New York, 1997, p. 27.
11. B. S. Mistry, T. Reineccius, and L. K. Olson, Gas chromatographyolfactometry
for the determination of key odorants in foods, Techniques for Analyzing Food
Aroma (R. Marsili, ed.), Marcel Dekker, Inc., New York, 1997, p. 265.
12. I. Blank, Gas chromatographyolfactometry in food aroma analysis, Techniques
for Analyzing Food Aroma (R. Marsili, ed.), Marcel Dekker, Inc., New York, 1997,
p. 293.
13. K. Coulibaly and I. J. Jeon, An overview of solid-phase extraction of food flavor
compounds and chemical residues, Food Rev. Int. 12(1):131 (1996).
14. R. E. Majors, The changing role of extraction in preparation of solid samples, LC-
GC Int. 10:638 (1996).
15. P. X. Etievant, Artifacts and contaminants in the analysis of food flavor, Crit. Rev.
Food Sci. Nutri. 36(7):733 (1996).
16. R. E. Majors, Liquid extraction techniques for sample preparation, LC-GC Int. 2:
93 (1997).
17. R. Kaiser, Environmental scents at the Lugurian Coast, Perfum. Flavor. 22:7
(1997).
18. B. E. Richter, B. A. Jones, J. L. Ezzell, and N. L. Porter, Accelerated solvent extrac-
tion: a technique for sample preparation, Anal. Chem. 68(6):1033 (1996).
19. A. Baudot, M. Marin, and H. E. Spinnler, The recovery of sulfur aroma compounds
of biological origin by pervaporation, Flavour Science. Recent Developments
(A. J. Taylor and D. S. Mottram, eds.), Hartnolls Ltd., Bodmin, UK, 1996, p. 301.
20. I. Souchon, C. Fontanini, and A. Voilley, Extraction of aroma compounds by
pervaporation, Flavour Science. Recent Developments (A. J. Taylor and D. S.
Mottram, eds.), Hartnolls Ltd., Bodmin, UK, 1996, p. 305.
21. C. C. Grimm, S. W. Lloyd, J. A. Miller, and A. M. Spanier, The analysis of food
volatiles using direct thermal desorption, Techniques for Analyzing Food Aroma
(R. Marsili, ed.), Marcel Dekker, Inc., New York, 1997, p. 59.
22. M. Rothaupt, Thermodesorption as sample preparation technique for food and fla-
vor analysis by gas chromatography, Flavor Analysis. Developments in Isolation
and Characterization (C. J. Mussinan and M. J. Morello, eds.), American Chemical
Society, Washington, DC, 1998, p. 116.
23. P. Werkhoff, M. Guntert, G. Krammer, H. Sommer, and J. Kaulen, Vacuum head-
space method in aroma research: flavor chemistry of yellow passion fruits, J. Agric.
Food Chem. 46(3):1076 (1998).
24. Th. McGee and K. Purzycki, Novel techniques for capturing scents in the rainforest
canopy: the final frontier, Perfum. Flavor. 24:1 (1999).
25. J. G. Wilkes, E. D. Conte, Y. Kim, M. Holcomb, J. B. Sutherland, and D. W.
Miller, Sample preparation for the analysis of flavors and off-flavors in foods, J.
Chromatogr. A 880:3 (2000).
26. C. Sarrazin, J. L. Le Quere, C. Gretsch, and R. Liardon, Representativeness of
coffee aroma extracts: a comparison of different extraction methods, Food Chem.
70:99 (2000).
196 Werkhoff et al.
27. K. Gassenmeier, Thermodesorption in flavour analysis, Frontiers of Flavour Sci-

ence (P. Schieberle and K. H. Engel, eds.), WB-Druck GmbH, Rieden am Forg-
gensee, 2000, p. 135.
28. R. P. Berladi and J. Pawliszyn, The application of chemically modified fused silica
fibres in the extraction of organics from water matrix samples and their rapid trans-
fer to capillary columns, Water Pollut. Res. J. Can. 24:179 (1989).
29. C. L. Arthur and J. Pawliszyn, Solid phase microextraction with thermal desorption
using fused silica optical fibers, Anal. Chem. 62:2145 (1990).
30. C. L. Arthur, D. Potter, K. Buchholz, S. Motlagh, and J. Pawliszyn, Solid phase
microextraction for the direct analysis of water: theory and practice, LC-GC 10:
656 (1992).
31. D. Louch, S. Motlagh, and J. Pawliszyn, Dynamics of organic compound extraction
from water using solid phase microextraction with fused silica fibers, Anal. Chem.
64:1187 (1992).
32. C. L. Arthur, L. Killam, K. Buchholz, and J. Pawliszyn, Automation and optimiza-
tion of solid phase microextraction, Anal. Chem. 64:1960 (1992).
33. Z. Zhang, M. J. Yang, and J. Pawliszyn, Solid phase microextraction, Anal. Chem.
66:844A (1994).
34. R. Eisert and K. Levsen, Solid-phase microextraction coupled to gas chromatogra-
phy: a new method for the analysis of organics in water, J. Chromatogr. A 733:
143 (1996).
35. X.-P. Lee, T. Kumazawa, K. Sato, and O. Suzuki, Detection of tricyclic antidepres-
sants in whole blood by headspace solid-phase microextraction and capillary gas
chromatography, J. Chromatogr. Sci. 35:302 (1997).
36. J. Pawliszyn, Solid Phase Microextraction. Theory and Practice, Wiley-VCH, New
York/Weinheim, 1997.
37. J. Pawliszyn (ed.), Applications of Solid Phase Microextraction, MPG Books Ltd.,
Bodmin, Cornwall, UK, 1999.
Food Chem. 42:1925 (1994).
39. F. Pelusio, T. Nilsson, L. Montanarella, R. Tilio, B. Larsen, S. Facchetti, and J.
Madsen, Headspace solid-phase microextraction analysis of volatile organic sulfur
compounds in black and white truffle aroma, J. Agric. Food Chem. 43:2138 (1995).
40. X. Yang and T. Peppard, Solid-phase microextraction of flavor compoundsa
comparison of two fiber coatings and a discussion of the rules of thumb for adsorp-
tion, LC-GC 13(11):882 (1995).
41. D. de la Calle Garcia, S. Magnaghi, M. Reichenbacher, and K. Danzer, Systematic
optimization of the analysis of wine bouquet components by solid-phase microex-
traction, J. High Resol. Chromatogr. 19:257 (1996).
42. A. J. Matich, D. D. Rowan, and N. H. Banks, Solid phase microextraction for
quantitative headspace sampling of apple volatiles, Anal. Chem. 68:4114 (1996).
43. A. Steffen and J. Pawliszyn, Analysis of flavor volatiles using headspace solid-
phase microextraction, J. Agric. Food Chem. 44:2187 (1996).
44. J. S. Elmore, M. A. Erbahadir, and D. S. Mottram, Comparison of dynamic head-
space concentration on Tenax with solid phase microextraction for the analysis of
aroma volatiles, J. Agric. Food Chem. 45:2638 (1997).
45. H. Verhoeven, T. Beuerle, and W. Schwab, Solid-phase microextraction: artefact

formation and its avoidance, Chromatographia 46:63 (1997).
46. A. D. Harmon, Solid-phase microextraction for the analysis of flavors, Techniques
for Analyzing Food Aroma (R. Marsili, ed.), Marcel Dekker, Inc., New York, 1997,
p. 81.
47. C. P. Bicchi, O. M. Panero, G. M. Pellegrino, and A. C. Vanni, Characterization
of roasted coffee and coffee beverages by solid phase microextractiongas chro-
matography and principal component analysis, J. Agric. Food Chem. 45:4680
(1997).
48. D. de la Calle Garcia, M. Reichenbacher, K. Danzer, Ch. Hurlbeck, Ch. Bartzsch,
and K.-H. Feller, Investigations on wine bouquet components by solid-phase mi-
croextractioncapillary gas chromatography (SPME-CGC) using different fibers,
J. High Resol. Chromatogr. 20:665 (1997).
49. D. Ulrich, A. Krumbein, and A. Rapp, Gaschromatographische Analyse der Aroma-
stoffe von Erdbeere, Kirsche und Tomate nach Festphasen-Mikroextraktion,
Deutsche Lebensmittel Rundschau 93(10):311 (1997).
50. G. Paliyath, M. D. Whiting, M. A. Stasiak, D. P. Murr, and B. S. Clegg, Volatile
production and fruit quality during development of superficial scald in Red Deli-
cious apples, Food Res. Int. 30(2):95 (1997).
51. J. Song, L. Fan, and R. M. Beaudry, Application of solid phase microextraction and
gas chromatography/time-of-flight mass spectrometry for rapid analysis of flavor
volatiles in tomato and strawberry fruits, J. Agric. Food Chem. 46(9):3721 (1998).
52. E. P. Jarvenpaa, Z. Zhang, R. Huopalahti, and J. W. King, Determination of fresh
onion (allium cepa) volatiles by solid phase microextraction combined with gas
chromatographymass spectrometry, Z. Lebensm. Unters. Forsch. A 207(1):39
(1998).
53. S. Francioli, M. Guerra, E. Lopez-Tamames, J. M. Guadayoi, and J. Caixach,
Aroma of sparkling wines by headspace/solid phase microextraction and gas
chromatography/mass spectrometry, Amer. J. Enology and Viticulture 50(4):404
(1999).
54. X. M. Wan, R. J. Stevenson, X. D. Chen, and L. D. Melton, Application of head-
space solid-phase microextraction to volatile flavour profile development during
storage and ripening of kiwifruit, Food Res. Int. 32(3):175 (1999).
55. M. E. Miller and J. D. Stuart, Comparison of gas-sampled and SPME-sampled
static headspace for the determination of volatile flavor components, Anal. Chem.
71:23 (1999).
56. H. Kataoka, H. L. Lord, and J. Pawliszyn, Applications of solid-phase microextrac-
tion in food analysis, J. Chromatogr. A 880:35 (2000).
57. F. Augusto, A. L. P. Valente, E. dos Santos Tada, and S. R. Rivellino, Screening
of Brazilian fruit aromas using solid-phase microextractiongas chromatography
mass spectrometry, J. Chromatogr. A 873:117 (2000).
58. M. Zhou, K. Robards, M. Glennie-Holmes, and S. Helliwell, Contribution of vola-
tiles to the flavour of oatmeal, J. Sci. Food Agric. 80(2):247 (2000).
59. R. T. Marsili and N. Miller, Determination of major aroma impact compounds in
fermented cucumbers by solid-phase microextractiongas chromatographymass
spectrometryolfactometry detection, J. Chromatogr. Sci. 38:307 (2000).
198 Werkhoff et al.
60. C. C. Grimm, S. W. Lloyd, R. Batista, and P. V. Zimba, Using microwave distilla-

tionsolid phase microextractiongas chromatographymass spectrometry for ana-
lyzing fish tissue, J. Chromatogr. Sci. 38:289 (2000).
61. E. Wiesenfeld, SPME/GC-Analyse von Duftkomponenten im Shampoo, Supelco
Reporter 17(1):7 (1998).
62. R. C. Galipo, A. J. Canhoto, M. D. Walla, and S. L. Morgan, Analysis of volatile
fragrance and flavor compounds by headspace solid phase microextraction and GC-
MS, J. Chem. Ed. 76(2):245 (1999).
63. T. Kaneko, Application of SPME in criminal investigations, Applications of Solid
Phase Microextraction (J. Pawliszyn, ed.), MPG Books Ltd., Bodmin, Cornwall,
UK, 1999, p. 573.
64. J. A. Field, G. Nickerson, D. D. James, and Ch. Heider, Determination of essential
oils in hops by headspace solid-phase microextraction, J. Agric. Food Chem. 44:
1768 (1996).
65. W. M. Coleman III and B. M. Lawrence, A comparison of selected analytical ap-
proaches to the analysis of an essential oil, Flav. Fragr. J. 12:1 (1997).
66. B. D. Mookherjee, S. M. Patel, R. W. Trenkle, and R. A. Wilson, A novel technol-
ogy to study the emission of fragrance from the skin, Perfum. Flavor. 23:1 (1998).
67. A. J. Matich, Analysis of food and plant volatiles, Applications of Solid Phase
Microextraction (J. Pawliszyn, ed.), MPG Books Ltd., Bodmin, Cornwall, UK,
1999, p. 349.
68. J. Rohloff, Monoterpene composition of essential oil from peppermint with regard
to leaf position using solid-phase microextraction and gas chromatography/mass
spectrometry analysis, J. Agric. Food Chem. 47(9):3782 (1999).
69. Th. J. Betts, Solid phase microextraction of volatile constituents from individual
fresh Eucalyptus leaves of three species, Planta Medica 66(2):193 (2000).
70. K.-H. Kubeczka, New approaches in essential oil analysis using polymer-coated
silica fibers, Proc. of the 29th International Symposium on Essential Oil, Frankfurt,
1998, p. 139.
71. M. Coleman III, T. A. Perfetti, and B. M. Lawrence, Automatic injection solid-
phase microextractionchiralgas chromatographymass selective detection anal-
yses of essential oils, J. Chromatogr. Sci. 36:575 (1998).
72. A. B. Caragay and A. D. Little, Supercritical fluids for extraction of flavors and
fragrances from natural products, Perfum. Flavor. 6:43 (1981).
73. W. Eisenbach, P. J. Gottsch, K. Niemann, and K. Zosel, Extraction with supercriti-
cal gases: the first twenty years, Fluid Phase Equilibria 10:315 (1983).
74. K.-W. Quirin, D. Gerard, and J. Kraus, Hochdruckextraktion mit Kohlendioxid,
Gordian 9:156 (1986).
75. O. G. Vitzthum, Extraktion mit uberkritischen Gasen, Taschenbuch fur Lebensmit-
telchemiker undtechnologen Band 2 (D. Osteroth, ed.), Springer-Verlag, Berlin,
Heidelberg, 1991, p. 61.
76. P. Pellerin, Supercritical fluid extraction of natural raw materials for the flavor and
perfume industry, Perfum. Flavor. 16:37 (1991).
77. J. Pawliszyn, Kinetic model of supercritical fluid extraction, J. Chromatogr. Sci.
31:31 (1993).
78. K. Kerrola, Literature Review: Isolation of essential oils and flavor compounds by
dense carbon dioxide, Food Rev. Int. 11(4):547 (1995).
79. W. K. Modey, D. A. Mulholland, and M. W. Raynor, Analytical supercritical fluid

extraction of natural products, Phytochem. Anal. 7:1 (1996).
80. G. M. Schneider, E. Stahl, and G. Wilke (eds.), Extraction with Supercritical Gases,
Verlag Chemie, Weinheim, 1980.
81. E. Stahl, K. W. Quirin, and D. Gerard, Verdichtete Gase zur Extraktion und Raffi-
nation, Springer-Verlag, Berlin, Heidelberg, 1987.
82. B. A. Charpentier and M. R. Sevenants (eds.), Supercritical Fluid Extraction and
Chromatography. Techniques and Applications, ACS Symposium Series 366,
American Chemical Society, Washington, DC, 1988.
83. M. B. King and T. R. Bott (eds.), Extraction of Natural Products Using Near-Criti-
cal Solvents, Blackie Academic & Professional, London, 1993.
84. S. S. H. Rizvi (ed.), Supercritical Fluid Processing of Food and Biomaterials,
Blackie Academic & Professional, London, 1994.
85. M. A. McHugh and V. J. Krukonis, Supercritical Fluid Extraction, Butterworth-
Heinemann, Boston, 1994.
86. M. A. Abraham and A. K. Sunol (eds.), Supercritical Fluids. Extraction and Pollu-
tion Prevention, ACS Symposium Series 670, American Chemical Society, Wash-
ington, DC, 1997.
87. A. Obaya-Valdivia and C. Guerrero-Barajas, Supercritical fluid extraction applied
to food processing industries, Tecnol. Aliment. 28:22 (1993).
88. S. Polesello, F. Lovati, A. Rizzolo, and C. Rovida, Supercritical fluid extraction
as a preparative tool for strawberry aroma analysis, J. High Resol. Chromatogr.
16:555 (1993).
89. D. L. Taylor and D. K. Larick, Volatile content and sensory attributes of supercriti-
cal carbon dioxide extracts of cooked chicken fat, J. Food Sci. 60(6):1197 (1995).
90. H. Kallio and K. Kerrola, Dense gas extraction as a preparation method in food
analysis, Curr. Status Future Trends Anal. Food Chem., Proc. Eur. Conf. Food
Chem., 8th, Volume 1 (G. Sontag and W. Pfannhauser, eds.), Austrian Chemical
Society, Vienna, 1995, p. 30.
91. O. Diaz, A. Cobos, L. de la Hoz, and J. A. Ordonez, Supercritical carbon dioxide
in the production of food from plants, Aliment. Equipos Tecnol. 16(8):55
(1997).
92. M. Valcarcel and M. T. Tena, Applications of supercritical fluid extraction in food
analysis, Fresenius J. Anal. Chem. 358(5):561 (1997).
93. D. R. Gere, L. G. Randall, and D. Callahan, Instrumental methods in food analysis.
Supercritical fluid extraction: principles and applications, Techn. Instrum. Anal.
Chem. 18:421 (1997).
94. E. Anklam, H. Berg, L. Mathiasson, M. Sharman, and F. Ulberth, Supercritical
fluid extraction (SFE) in food analysis: a review, Food Addit. Contam. 15(6):729
(1998).
95. D. A. Moyler, Extraction of essential oils with carbon dioxide, Flav. Fragr. J. 8:
235 (1993).
96. P. Castioni, P. Christen, and J. L. Veuthey, Supercritical fluid extraction of com-
pounds from plant origin, Analusis 23(3):95 (1995).
97. R. M. Smith, Supercritical fluid extraction of natural products, LC-GC Int. 9(1):8
(1996).
98. J. A. Pino, J. Garcia, and M. A. Martinez, Comparative chemical composition of
200 Werkhoff et al.
the volatiles of Coleus aromaticus produced by steam distillation, solvent extraction

and supercritical carbon dioxide extraction, J. Essent. Oil Res. 8:373 (1996).
99. M. H. Boelens and H. Boelens, Differences in chemical and sensory properties of
orange flower and rose oils obtained from hydrodistillation and from supercritical
CO 2 extraction, Perfum. Flavor. 22:31 (1997).
100. E. Reverchon, G. Della Porta, and D. Gorgolione, Supercritical CO 2 extraction of
volatile oil from rose concrete, Flav. Frag. J. 12:37 (1997).
101. J. Garcia-Fajardo, M. Martinez-Sosa, M. Estarron-Espinosa, G. Vilarem, A. Gaset,
and J. M. de Santos, Comparative study of the oil and supercritical CO 2 extract of
Mexican Pimento (Pimenta dioica Merill), J. Essent. Oil Res. 9:181 (1997).
102. A. P. Jarvis and E. D. Morgan, Isolation of plant products by supercritical-fluid
extraction, Phytochem. Anal. 8:217 (1997).
103. H. Surburg, M. Guntert, and H. Harder, Investigations of volatiles from flowers.
Analytical and olfactory aspects, Recent Developments in Flavor and Fragrance
Chemistry (R. Hopp and K. Mori, eds.), VCH Verlagsgesellschaft, Weinheim,
1993, p. 103.
104. W. Engel, W. Bahr, and P. Schieberle, Solvent assisted flavour evaporationa
new and versatile technique for the careful and direct isolation of aroma compounds
from complex food matrices, Eur. Food Res. Technol. 209:237 (1999).
105. S. J. MacDonald and D. Wheeler, Fast GC temperature programmingPart 2: Re-
sistive heating with conventional GCs, Int. Lab. News 20C (1998).
106. M. van Deursen, J. Beens, and C. A. Cramers, Possibilities and limitations of fast
temperature programming as a route towards fast GC, J. High Resol. Chromatogr.
22(9):509 (1999).
107. J. Dalluge, R. Ou-Aissa, J. J. Vreuls, and U. A. Th. Brinkmann, Fast temperature
programming in gas chromatography using resistive heating, J. High Resol. Chro-
matogr. 22(8):459 (1999).
108. A. Grall, C. Leonard, and R. Sacks, Peak capacity, peak-capacity production rate,
and boiling point resolution for temperatureprogrammed GC with very high pro-
gramming rates, Anal. Chem. 72(3):591 (2000).
109. E. Baltussen, P. Sandra, F. David, and C. A. Cramers, Stir Bar Sorptive Extraction
(SBSE), a novel extraction technique for aqueous samples: theory and principles,
J. Microcolumn Sep. 11:737 (1999).
110. A. G. J. Tredoux, H. H. Lauer, Th. Heidemann, and P. Sandra, Stir bar sorptive
extraction applied to lemon flavoured beverages, 23rd International Symposium on
Capillary Chromatography, CD-ROM, M.38, Riva del Garda, 2000.
111. B. D. Mookherjee and R. W. Trenkle, Sun-scents: The production of aroma mole-
cules by the action of sunlight on the essential oil of cotton, Proc. 11th Intern. Congr.
Essent. Oils, Fragrances and Flavours, New Delhi, Vol. 2, 1989, p. 69.
112. W. Grosch, Neuere Entwicklungen in der Aromaforschung, Lebensmittelchemie
47:129 (1993).
113. W. Grosch, Detection of potent odorants in foods by aroma extract dilution analysis,
Trends Food Sci. Technol. 4:68 (1993).
114. W. Grosch, Determination of potent odourants in food by aroma extract dilution
analysis (AEDA) and calculation of odour activity values (OAVs), Flav. Fragr. J.
9:147 (1994).
115. W. Grosch, Ergebnisse aus der Aromaforschung, Gordian 11 :171 (1997).

116. W. Grosch, Welche Verbindungen bevorzugt der Geruchssinn bei erhitzten Lebens-
mitteln? Lebensmittelchemie 52:143 (1998).
117. E. Guichard, Quantification of some volatile aromatic compounds of apricot by
adding standards, J. Food Sci. 53(6):1902 (1988).
118. M. Preininger, Quantitation of potent food aroma compounds by using stable iso-
tope labeled and unlabeled internal standard methods, Food Flavors: Formation,
Analysis and Packaging Influences (E. T. Contis, C.-T. Ho, C. J. Mussinan, T. H.
Parliment, F. Shahidi, and A. M. Spanier, eds.), Elsevier Science B. V., Amsterdam,
1998, p. 87.
119. W. Schmidt and W. Grosch; Quantitative Analyse fluchtiger Aromastoffe mit hohen
Aromawerten in Sauerkirschen (Prunus cerasus L.), Sukirschen (Prunus avium
L.) und Kirschkonfituren, Z. Lebensm. Unters. Forsch. 183:39 (1986).
120. P. Schieberle and W. Grosch, Quantitative analysis of important volatile flavour
compounds in fresh and stored lemon oil/citric acid emulsions, Lebensm. Wiss.
Technol. 21 :158 (1988).
121. P. Schieberle and W. Grosch, Potent odorants resulting from the peroxidation of
lemon oil, Z. Lebensm. Unters. Forsch. 189:26 (1989).
122. R. G. Buttery, D. J. Stern, and L. C. Ling, Studies on flavor volatiles of some sweet
corn products, J. Agric. Food Chem. 42:791 (1994).
123. R. G. Buttery, L. C. Ling, and D. J. Stern, Studies on popcorn aroma and flavor
volatiles, J. Agric. Food Chem. 45:837 (1997).
124. L. B. Fay, S. Metairon, J. Lin, and I. Blank, Stable isotope dilution assay mass
spectrometry in flavour research: internal standard and calibration issues, Frontiers
of Flavour Science (P. Schieberle and K. H. Engel, eds.), WB-Druck GmbH, Rieden
am Forggensee, 2000, p. 125.
125. O. Hahn, Uber das Uran Z und seine Muttersubstanz, Z. Phys. Chem. 103:461
(1923).
126. G. v. Hevesy and R. Hobbie, Uber die Ermittlung des Bleigehaltes von Gesteinen,
Z. Anal. Chem. 88:1 (1932).
127. R. Schoenheimer, S. Ratner, and D. Rittenberg, J. Biol. Chem. 130:703 (1939).
128. H. H. Ussing, Analysis of protein by means of deuterium-containing amino-acids,
Nature 144:977 (1939).
129. D. Rittenberg and G. L. Foster, A new procedure for quantitative analysis by isotope
dilution, with application to the determination of amino acids and fatty acids, J.
Biol. Chem. 133:737 (1940).
130. Ch. C. Sweeley, W. H. Elliott, I. Fries, and R. Ryhage, Mass spectrometric determi-
nation of unresolved components in gas chromatographic effluents, Anal. Chem.
38(11):1549 (1966).
131. I. Bjorkhem, R. Blomstrand, O. Lantto, L. Svensson, and G. Ohman, Toward abso-
lute methods in clinical chemistry: application of mass fragmentography to high-
accuracy analyses, Clin. Chem. 22(11):1789 (1976).
132. W. K. Rohwedder, Mass spectrometry of lipids labeled with stable isotopes, Prog.
Lipid Res. 24:1 (1985).
133. F. B. Suhre, R. M. Simpson, and J. W. Shafer, Qualitative/quantitative determina-
tion of sulfamethazine in swine tissue by gas chromatographic/electron impact
202 Werkhoff et al.
mass spectrometry using a stable isotope labeled internal standard, J. Agric. Food
Chem. 29:727 (1981).
134. J. R. Startin, I. Parker, M. Sharman, and J. Gilbert, Analysis of di-(2-ethylhexyl)adi-
pate plasticiser in foods by stable isotope dilution gas chromatographymass spec-
trometry, J. Chromatogr. 387:509 (1987).
135. L. Castle, A. J. Mercer, J. R. Startin, and J. Gilbert, Migration from plasticized
films into foods 3. Migration of phthalate, sebacate, citrate and phosphate esters
from films used for retail food packaging, Food Addit. Contam. 5(1):9 (1988).
136. K. Y. Patterson, C. Veillon, Ph. B. Moser-Veillon, and G. F. Wallace, Determina-
tion of zinc stable isotopes in biological materials using isotope dilution inductively
coupled plasma mass spectrometry, Anal. Chim. Acta 258:317 (1992).
137. M. Rychlik and P. Schieberle, Quantification of the mycotoxin patulin by a stable
isotope dilution assay, J. Agric. Food Chem. 47:3749 (1999).
138. M. Rychlik, Quantification of free and bound pantothenic acid in foods and
blood plasma by a stable isotope dilution assay, J. Agric. Food Chem. 48:1175
(2000).
139. Th. J. Evans, Ch. E. Butzke, and S. E. Ebeler, Analysis of 2,4,6-trichloroanisole
in wines using solid-phase microextraction coupled to gas chromatographymass
spectrometry, J. Chromatogr. A 786:293 (1997).
140. J. F. Pickup and K. McPherson, Theoretical considerations in stable isotope dilution
mass spectrometry for organic analysis, Anal. Chem. 48(13):1885 (1976).
141. E. D. Bush and W. F. Trager, Analysis of linear approaches to quantitative stable
isotope methodology in mass spectrometry, Biomed. Mass Spectrom. 8(5):211
(1981).
142. K. G. Heumann, Isotope dilution mass spectrometry of inorganic and organic sub-
stances, Fresenius Z. Anal. Chem. 325:661 (1986).
143. P. De Bievre, Isotope dilution mass spectrometry: what can it contribute to accuracy
in trace analysis? Fresenius J. Anal. Chem. 337:766 (1990).
144. J. F. Sabot, Concepts et equivoques en dilution isotopique stable. Considerations
mathematiques fondamentales, Analusis 22:381 (1994).
145. J. R. Troost and E. Y. Olavesen, Gas chromatographic/mass spectrometric calibra-
tion bias, Anal. Chem. 68:708 (1996).
146. L. M. Thienpont, B. van Nieuwenhove, D. Stockl, and A. P. de Leenheer, Calibra-
tion for isotope dilution mass spectrometrydescription of an alternative to the
bracketing procedure, J. Mass Spectrom. 31 :1119 (1996).
147. R. L. N. Harris, M. J. Lacey, W. V. Brown, and M. S. Allen, Determination of 2-
methoxy-3-alkylpyrazines in wine by gas chromatography/mass spectrometry,
Vitis 26:201 (1987).
148. P. Schieberle and W. Grosch, Quantitative analysis of aroma compounds in wheat
and rye bread crusts using a stable isotope dilution assay, J. Agric. Food Chem.
35:252 (1987).
149. B. Dollmann, D. Wichmann, A. Schmitt, H. Koehler, and P. Schreier, Quantitative
analysis of 2-aminoacetophenone in off-flavored wines by stable isotope dilution
assay, J. AOAC Int. 79(2):583 (1996).
150. V. Aubry, P. X. Etievant, Ch. Ginies, and R. Henry, Quantitative determination of
potent flavor compounds in Burgundy Pinot Noir wines using a stable isotope dilu-
tion assay, J. Agric. Food Chem. 45:2120 (1997).
151. H. Guth, Quantitation and sensory studies of character impact odorants of different
white wine varieties, J. Agric. Food Chem. 45:3027 (1997).
152. C. Milo and G. A. Reineccius, Identification and quantification of potent odorants
in regular-fat and low-fat mild Cheddar cheese, J. Agric. Food Chem. 45:3590 (1997).
153. I. Blank, L. B. Fay, F. J. Lakner, and M. Schlosser, Determination of 4-hydroxy-
2,5-dimethyl-3(2H)-furanone and 2(or 5)-ethyl-4-hydroxy-5(or 2)-methyl-3(2H)-
furanone in pentose sugar-based Maillard model systems by isotope dilution assays,
J. Agric. Food Chem. 45:2642 (1997).
154. M. S. Allen and M. J. Lacey, Stable isotope dilution gas chromatographymass
spectrometry for determination of methoxypyrazines (green aroma) in wine,
Mod. Methods Plant Anal. 19:193 (1997).
155. Y. Kotseridis, R. L. Baumes, A. Bertrand, and G. K. Skouroumounis, Quantitative
determination of 2-methoxy-3-isobutylpyrazine in red wines and grapes of Bor-
deaux using a stable isotope dilution assay, J. Chromatogr. A 841:229 (1999).
156. Y. Kotseridis, R. L. Baumes, and G. K. Skouroumounis, Quantitative determination
of free and hydrolytically liberated -damascenone in red grapes and wines using
a stable isotope dilution assay, J. Chromatogr. A 849:245 (1999).
157. Y. Kotseridis, R. L. Baumes, A. Bertrand, and G. K. Skouroumounis, Quantitative
determination of -ionone in red wines and grapes of Bordeaux using a stable iso-
tope dilution assay, J. Chromatogr. A 848:317 (1999).
158. A. P. Pollnitz, G. P. Jones, and M. A. Sefton, Determination of oak lactones in
barrel-aged wines and in oak extracts by stable isotope dilution assay, J. Chro-
matogr. A 857:239 (1999).
159. A. P. Pollnitz, K. H. Pardon, and M. A. Sefton, Quantitative analysis of 4-ethyl-
phenol and 4-ethylguaiacol in red wine, J. Chromatogr. A 874:101 (2000).
160. P. Schieberle, New developments in methods for analysis of volatile flavor com-
pounds and their precursors, Characterization of Food. Emerging Methods (A. G.
Gaonkar, ed.), Elsevier Science B. V., Amsterdam, 1995, p. 403.
161. Ch. Milo and I. Blank, Quantification of impact odorants in food by isotope dilution
assay: strength and limitations, Flavor Analysis. Developments in Isolation and
Characterization (C. J. Mussinan and M. J. Morello, eds.), ACS Symposium Series
705, American Chemical Society, Washington, DC, 1998, p. 250.
162. I. Blank, Ch. Milo, J. Lin, and L. B. Fay, Quantification of aroma-impact compo-
nents by isotope dilution assayRecent developments, Flavor Chemistry. Thirty
Years of Progress (R. Teranishi, E. L. Wick, and I. Hornstein, eds.), Kluwer
Academic/Plenum Publishers, New York, 1999, p. 63.
163. A Sen and W. Grosch, Synthesis of six deuterated sulfur containing odorants to
be used as internal standards in quantification assays, Z. Lebensm. Unters. Forsch.
192:541 (1991).
164. A. Sen, P. Schieberle, and W. Grosch, Quantitative determination of 2,5-dimethyl-
4-hydroxy-3(2H)-furanone and its methyl ether using a stable isotope dilution
assay, Lebensm. Wiss. Technol. 24:364 (1991).
165. H. Rakoff and W. K. Rohwedder, Catalytic deuteration of alkynols and their tetra-
hydropyranyl ethers, Lipids 27(7):567 (1992).
166. C. Masanetz, I. Blank, and W. Grosch, Synthesis of [2H 6]-3-mercapto-3-methylbu-
tyl formate to be used as internal standard in quantification assays, Flav. Fragr. J.
10:9 (1995).
204 Werkhoff et al.
167. Y. Kotseridis, R. Baumes, and G. K. Skouroumounis, Synthesis of labeled [2H 4]-

damascenone, [2H 2]2-methoxy-3-isobutylpyrazine, [2H 3]-ionone, and [2H 3]-io-
none for quantification in grapes, juices and wines, J. Chromatogr. A 824:71 (1998).
168. J. Lin, D. H. Welti, F. A. Vera, L. B. Fay, and I. Blank, Synthesis of deuterated
volatile lipid degradation products to be used as internal standards in isotope dilu-
tion assays. 1. Aldehydes, J. Agric. Food Chem. 47:2813 (1999).
169. J. Lin, D. H. Welti, F. A. Vera, L. B. Fay, and I. Blank, Synthesis of deuterated
volatile lipid degradation products to be used as internal standards in isotope dilu-
tion assays. 2. Vinyl Ketones, J. Agric. Food Chem. 47:2822 (1999).
170. S. Fuchs, Th. Beck, S. Burkardt, M. Sandvoss, and A. Mosandl, Biogenetic studies
in Mentha x piperita. 1. Deuterium-labeled monoterpene ketones: Synthesis and
stereoselective analysis, J. Agric. Food Chem. 47:3053 (1999).
171. J. Lin, L. B. Fay, D. H. Welti, and I. Blank, Synthesis of trans-4,5-epoxy-(E)-2-
decenal and its deuterated analog used for the development of a sensitive and selec-
tive quantification method based on isotope dilution assay with negative chemical
ionization, Lipids 34(10):1117 (1999).
172. A. Buettner and P. Schieberle, Stable isotope dilution assays for the quantification
of odour-active thiols in hand-squeezed grapefruit juices (Citrus Paradisi Mac-
Fayden), Frontiers of Flavour Science (P. Schieberle and K. H. Engel, eds.), WB-
Druck GmbH, Rieden am Forggensee, 2000, p. 132.
173. D. D. Roberts, P. Pollien, and C. Milo, Solid-phase microextraction method devel-
opment for headspace analysis of volatile flavor compounds, J. Agric. Food Chem.
48:2430 (2000).
7
SPME Comparison Studies and
What They Reveal
Ray Marsili
Dean Foods Company, Rockford, Illinois
I. INTRODUCTION
Solid-phase microextraction (SPME) appears to offer flavor and fragrance chem-

ists significant benefits as an extraction/concentration technique prior to gas chro-
matography (GC) analysis. It is simple, rapid, solventless, and sensitive. Further-
more, it is less expensive to implement than purge-and-trap methods and many
other sample preparation techniques that require specialized ancillary instrumen-
tation. Before adapting a new extraction technology like SPME as a routine pro-
cedure, it is important to understand how it compares to the standard sample
preparation techniques that have been used in the past.
Why consider SPME as an extraction technique for flavor, fragrance, and
odor studies? What, if any, advantages does SPME offer over other analyte
extraction/concentration techniques? Is SPME as accurate and sensitive as other
commonly used extraction techniques (e.g., purge-and-trap/dynamic headspace)?
Is it a preferred technique for some types of analytes and/or some types of sam-
ples? Are there some types of analytes and/or sample matrixes that should not
be analyzed by SPME? What are the weaknesses and biases of SPME-based
methods? Are there significant differences in chromatograms obtained by SPME
compared to other techniques? If so, does it mean SPME is an inferior technique
that should be avoided? How does SPME fiber type affect analytical results?
This chapter attempts to answer these questions by presenting application
examples that compare SPME results with results obtained by alternate standard
extraction methods and by examining examples that compare results obtained
with different fiber types. By considering important trends in these studies, it is
205
206 Marsili
possible to gain a better understanding of the strengths and weaknesses of SPME

as an analytical sample preparation tool for GC/MS analysis of flavors and fra-
grances. Also, these examples provide insights into SPME method optimization.
II. COMPARING SPME TO CLASSICAL SAMPLE

PREPARATION METHODS
A. Studies with PDMS Fibers
1. Ground Coffee, Fruit Juice, and Butter-Flavored
Vegetable Oil
Yang and Peppard (1) studied the recovery of flavor chemicals in espresso-roast
ground coffee, fruit juice beverage, and butter-flavored vegetable oil, comparing
SPME (100-m PDMS fibers in all cases) with other sample preparation methods.
a. Solid Espresso-Ground Coffee

SPME proved useful for headspace sampling of solid espresso ground coffee,
providing results similar to static headspace testing. For both static headspace
(HS) and SPME, ground coffee samples were heated to 120C. As evidenced in
the chromatograms shown in Fig. 1, static HS was more sensitive for highly
volatile compounds, whereas the SPME HS method detected more of the less
volatile compounds. One advantage that SPME offers over static HS is that it
prevents water from entering the GC capillary column and, therefore, avoids
potential column degradation.
b. Fruit Juice Beverages

A fruit juice beverage was analyzed by GC/MS following both SPME liquid
sampling and solvent extraction with dichloromethane (DCM). With the DCM
method, 250 mL of juice beverage was extracted three times with 50 mL DCM;
the solvent was then removed using a Kuderna-Danish evaporator and concen-
trated to 250 L with a gentle stream of nitrogen. For the liquid sampling SPME
technique, the 100-m PDMS fiber was inserted into a 4-mL vial containing 3.0
mL of juice beverage and 0.6 g of NaCl. After SPME liquid sampling for 10
minutes at ambient temperature, the SPME fiber was introduced into the GC
injector in splitless mode. The resulting chromatograms appear in Fig. 2.
Most of the flavor components extracted by DCM were also concentrated
on the SPME fiber, although at somewhat different relative recoveries. The sensi-
tivity of SPME achieved in this example was comparable to or higher than that
of the DCM-extraction method for most esters, terpenoids, and -decalactone.
Extraction recoveries of fatty acids by SPME with the PDMS fiber was poor.
SPME Comparison Studies 207
FIGURE 1 GC/MS chromatograms of espresso-roast ground coffee by (A) SPME

(PDMS) headspace sampling at 120C and (B) 1-mL conventional headspace injection at
120C. Peak identification is as follows: (1) pyridine; (2) 2-methylpyrazine; (3) acetol;
(4) acetic acid; (5) hydroxyacetone acetate; (6) furfuryl acetate; (7) 5-methylfurfural;
(8) -butyrolactone; (9) furfuryl alcohol; (10) maltol; (11) 2-acetylpyrrole; (12) unknown;
(13) 4-vinylguaiacol; (14) 3-hydroxypyridine.
However, the low affinity of PDMS SPME for fatty acids can be advantageous
for example, by reducing interference from these compounds when they might
interfere/coelute with other trace flavor constituents. (Note: It is possible to obtain
ppb level extraction of fatty acids with other types of SPME fibers as discussed
later in this chapter.)
c. Butter-Flavored Vegetable Oil

HS sampling by SPME works well for the analysis of vegetable oilbased sam-
ples. For SPME HS sampling, sensitivity is dependent on the sample matrix. In
this case, the sensitivity of SPME for flavor compounds in vegetable oilbased
samples is much less than in aqueous samples because the solubility of flavor
compounds is generally much greater in vegetable oil than in water. To compen-
208 Marsili
FIGURE 2 GC/MS chromatograms of fruit juice beverage by (A) DCM extraction and
(B) SPME (PDMS) liquid sampling. Peak identification is as follows: (1) dichloromethane;
(2) ethyl butyrate; (3) ethyl isovalerate; (4) limonene; (5) ethyl hexanoate; (6) isoamyl
butyrate; (7) hexanyl acetate; (8) cis-3-hexenyl acetate; (9) hexanol; (10) cis-3-hexenol;
(11) cis-3-hexenyl butyrate; (12) furfural; (13) benzaldehyde; (14) linalool; (15) -terpin-
eol; (16) butyric acid; (17) 2-methyl butyric acid; (18) -terpineol; (19) hexanoic acid;
(20) cis-methyl cinnamate; (21)1-(2-furyl)-2-hydroxyethanone; (22) furaneol; (23) trans-
methyl cinnamate; (24) -decalactone; (25) dodecanoic acid; (26) (hydroxymethyl)fur-
fural.
sate for this, HS extraction is conducted at elevated temperatures. SPME sam-

pling was able to detect diacetyl, -decalactone, and -dodecalactone in butter-
flavored vegetable oil. Most other sample preparation techniques, including static
HS sampling, simultaneous distillation/extraction, and purge-and-trap sampling,
would be less sensitive and/or substantially more time-consuming.
Based on these studies, Yang and Peppard conclude that SPME provides
many advantages over conventional sample preparation techniques. It can be
readily applied to flavor analysis of solid, liquid, and gaseous samples, and it
is especially suitable as a rapid screening technique. However, they warn that
quantitative results obtained with SPME are highly dependent on experimental
conditions and sample matrix and that external calibration techniques are not
reliable for quantitation because a synthetic matrix is usually not a good match
to an authentic sample.
2. Truffles
Pelusio et al. (2) studied volatiles, including organic sulfur compounds, in black
and white truffles, comparing results with SPME extraction to a purge-and-trap
sample preparation technique using Tenax as the adsorbent.
Immediately prior to analysis, the truffles were rinsed with tap water,
brushed, and air-dried. SPME analysis was performed with approximately 5 g
of freshly cut pieces of truffle flesh in a 25-mL glass vial closed with a Teflon-
lined membrane cap. For the headspace Tenax adsorption, thin strips of truffle
flesh (approximately 200 mg) were used.
In the SPME extraction method, the fiber was exposed to the headspace
above the truffle sample for 30 min at either 80C or at room temperature. In
the HS Tenax adsorption procedure, thin strips of sample material were placed
between plugs of glass wool in an empty stainless steel tube, which was then
mounted in a desorber oven. The oven was heated to 60C, and approximately
20 mL of He gas was used to flush the tube to the Tenax-filled cold (40C)
trap over a 4-min period. The trapped volatiles were desorbed by rapid electrical
heating to 250C for 45 seconds and transferred to a GC column.
Table 1 lists volatiles identified in white and black truffle aromas by head-
space SPME (100-m PDMS) GC/MS, and Table 2 lists results by purge-and-
trap (Tenax) GC/MS. Results obtained by HS-SPME-GC/MS agreed well with
those obtained by headspace Tenax adsorption GC/MS for the volatile organic
sulfur compounds, and the expected discrimination of the polar or very volatile
compounds by HS-SPME was confirmed. Pelusio et al. concluded that HS-
SPME-GC/MS is a powerful technique for analysis of volatile organic sulfur
compounds in truffle aromas, but because HS-SPME (with PDMS fibers) strongly
discriminates more polar and very volatile compounds, it is less suited for quanti-
tative analysis.
3. Tomatoes
Krumbein and Ulrich compared three sample preparation techniques for the deter-
mination of fresh tomato aroma volatiles (3). These researchers attempted to eval-
uate three sample preparation techniques prior to GC/MS analysis for their ability
to distinguish between two types of tomato cultivars.
210 Marsili
TABLE 1 Volatiles Identified in White and Black Truffle Aromas by

HS-SPME GC/MS
Area percentage
Compound White truffle Black truffle
dimethyl sulfide 3.6 6.5
2-butanone 81 78
2-butanol 3.8 14
dimethyl disulfide 2.0 NDa
bis(methylthio)methane 5.9 0.9
dimethyl trisulfide 2.7 ND
1,2,4-trithiolane 0.1 ND
methyl(methylthio)methyl disulfide 1.1 ND
tris(methylthio)methane 0.3 ND
a
ND not detected.
The methods used were as follows:

1. Dynamic headspace using Tenax TA. A sample of 500 g of tomato
plus 500 mL of saturated calcium chloride solution and an internal
standard (2-octanone) were mixed in a blender. The mixture was placed
in a 3-L flask. The flask was purged with purified air while the tomato
mixture was stirred with a magnetic stirrer; the tomato volatiles in the
air stream were passed through a 200-mg Tenax trap. The isolation
was carried out for 150 min, then the trap was removed and the vola-
tiles extracted with 3 mL of acetone. Samples were concentrated under
nitrogen to a volume of 50 L.
2. Headspace SPME. A blended mixture of 500 g tomatoes and 500 mL
of saturated sodium chloride solution was centrifuged at 5000 rpm for
30 minutes at 4C. An internal standard (2-octanone) was added to 12
mL of the supernatant, and the solution was transferred to a 20 mL
headspace vial. Volatiles from the headspace were adsorbed for 10 min
on a 100-m PDMS fiber at 30C.
3. Liquid-liquid extraction. Tomato fruits were blended and centrifuged
in a procedure similar to that used with the SPME technique. Volatiles
in the supernatant were extracted with Freon for 20 hours at room
temperature, and the extract concentrated to 10 L in a water bath at
23C using a Vigreux distillation column.
Fifty-four volatiles were identified by GC/MS in the dynamic headspace
extract. Only a small number of these are critical to tomato flavor. Therefore,
TABLE 2 Volatiles Identified in White and Black Truffle Aromas by

Headspace Tenax Adsorption GC/MS
Area percentage
Compound White truffle Black truffle
acetaldehyde 6.5 NDa
ethanol 45.0 1.1
acetone 2-propanol 7.9 8.5
dimethyl sulfide 18.9 3.8
1-propanol 3.2 ND
2,3-butanedione 0.1 ND
2-butanone 1.1 38.2
2-butanol 0.4 42.7
ethyl acetate 0.1 ND
2-methyl-1-propanol 1.4 ND
3-methylbutanal 0.4 TRb
2-methylbutanal TR TR
1-(methylthio)propane ND TR
1-(methylthio)-1-propene ND TR
3-methyl-1-butanol 0.1 ND
dimethyl disulfide 0.1 TR
2-methyl-1-butanol 1.2 TR
acetoin 0.7 ND
hexanal 0.2 ND
bis(methylthio)methane 8.8 ND
dimethyl trisulfide TR TR
1-octen-3-one ND 0.3
1-octen-3-olc 1.3 4.3
3-octanonec 0.3 0.7
3-octanolc 0.2 0.3
2-octen-1-ol ND TR
a
ND not detected.
b
TR trace.
c
Characteristic aroma of mushroom odor.
comparison of the isolation methods concentrated on 15 flavor compounds that

best characterize fresh tomato flavor based on their odor unit values. These com-
ponents include six aldehydes, three alcohols, and four ketones.
The chromatograms of tomato volatiles obtained by the three methods are
shown in Fig. 3. SPME was unable to detect the highly volatile flavor components
1-penten-3-one and 3-methyl butanol. All important aroma volatiles were de-
tected by liquid-liquid extraction. Phenylacetaldehyde and 2-phenylethanol were
212 Marsili
FIGURE 3 GC/MS chromatograms of tomato volatiles obtained by three different sample

preparation techniques: (A) dynamic headspace on Tenax; (B) headspace SPME (PDMS);
and (C) Freon liquid-liquid extraction. Peak identification is as follows: (1) 3-methylbuta-
nal; (2) 1-penten-3-one; (3) hexanal; (4) (Z)-3-hexenal; (5) 2-methyl butanol; (6) (E)-2-
hexenal; (7) (E)-2-heptenal; (8) 6-methyl-5-hepten-2-one; (9) (Z)-3-hexenol; (10) 2-isobu-
tylthiazole; (11) phenylacetaldehyde; (12) methyl salicylate; (13) geranylacetone; (14) 2-
phenylethanol; (15) -ionone; (IS) 2-octanone internal standard.
not measured quantitatively with either the dynamic headspace or the headspace-
SPME method.
The precision data for all three methods were comparable; the relative stan-
dard deviation for all components was less than 15%. With respect to sample
preparation time, the liquid extraction method was much more time-consuming,
requiring a sample preparation time of 1 hour and an extraction time of 20 hours.
The headspace analysis on a Tenax trap required a sample preparation time of
one hour and an extraction time of 2.5 hours. The most rapid method was the
HS-SPME method, with a sample preparation time of 10 minutes and an extrac-
tion time of 10 minutes.
All three sample preparation techniques were suitable for characterizing
the two cultivars. With all three methods, the relative peak areas of the aroma
volatiles hexanal and 2-isobutylthiazole differed significantly for the two cultivars
studied. Krumbein and Ulrich concluded that SPME is a suitable screening
method for tomato flavor.
4. Conclusions
SPME results that have been discussed so far, while covering a variety of food
types and aroma-contributing compounds, reveal a common theme: SPME ex-
traction methods with PDMS fibers exhibit a propensity for low recoveries with
polar and highly volatile organic compounds. All researchers agreed, however,
that SPME was, in most cases, as sensitive and precise as other techniques when
applied to nonpolar, less volatile, and semivolatile compounds. Also, all agreed
that SPME offers significant advantages with respect to ease of use and analysis
time.
It is important to note that the studies discussed above were conducted
between 1994 and 1996. PDMS SPME fibers were the predominant type of fiber
used prior to 1996. Since then, numerous additional fiber types have been devel-
oped and are significantly superior to PDMS for extracting polar organics and
highly volatile compounds.
It is not uncommon to hear some flavor chemists report that SPME doesnt
work as well as other extraction/concentration techniques. Unfortunately, many
havent tried the newer fibers and are basing their assessment of SPME on previ-
ous work done only with PDMS. Examples of successful aroma analyses of polar
and/or highly volatile analytes using some of these newer types of fibers follow.
B. Studies with Fibers Other Than PDMS

1. Milk
Marsili compared SPME and dynamic headspace (DH) GC/MS techniques for
the analysis of light-induced lipid oxidation products in milk (4). In the SPME
method, 3 g of milk (2% milkfat) and 4-methyl-2-pentanone internal standard
(10 L of a 20 ppm solution in methanol) were placed in a 9-mL vial and capped.
A 75-m Carboxen-1006/PDMS fiber was inserted into the headspace above the
milk sample. The Carboxen/PDMS fiber has a combination of micro-, meso-,
and macro-pores ranging from 6 to 50A. The volatile flavor compounds that are
the best indicators of light-induced oxidation in milk are pentanal, hexanal, and
dimethyl disulfide. The Carboxen/PDMS fiber was selected for this study because
it is well suited for the analysis of low-molecular-weight volatiles. Adsorption
of volatiles from the milk onto the SPME fiber was conducted at 45C for 15
min with stirring. The sealed vial was allowed to equilibrate for 2 min at 45C
before the SPME fiber was inserted.
With the purge-and-trap DH-GC/MS method, 20.0 g of milk and 4-methyl-
2-pentanone internal standard (10 L of a 20 ppm solution) were added to a 30-
mL impinger flask. The volatiles were trapped on a Tenax trap and then thermally
desorbed onto the front of the analytical column where they were refocused with
cryogenic cooling (100C) prior to injection into the analytical capillary
column.
Spiking milk samples with various levels of pentanal and hexanal revealed
that SPME with Carboxen/PDMS fibers was a better test for quantitating pentanal
and hexanal in milk than DH-GC/MS; SPME demonstrated superior precision
214 Marsili
without a sacrifice in sensitivity. Furthermore, fewer problems with carryover,

background, and artifact peaks were observed. In addition, SPME was signifi-
cantly faster. In both procedures, only trace chromatographic peaks were ob-
served for dimethyl disulfide in light-abused milks.
Table 3 shows the principal analytical parameters for pentanal and hexanal
in standard calibration samples analyzed by DH-GC and SPME-GC. Detection
limits were determined by analyzing decreasing concentrations of pentanal and
hexanal in 2% milk and skim milk. The lowest concentration was established as
that for which all the ions selected for a given compound could be differentiated
from the background. Repeatability was assessed by calculating the coefficient
of variance obtained by analyzing in quadruplicate the same sample of either 2%
milk or skim milk spiked with 2 ng/mL of pentanal and hexanal. The linear
correlation coefficients of the standard calibration curves were used as an estimate
to compare the accuracy of each technique for each analyte in 2% milk and skim
milk. This work showed that for the analysis of pentanal and hexanal in milk
samples, SPME analysis consistently demonstrated superior precision and equiv-
alent sensitivity compared to DH-GC.
2. Pickle Brine
Testing for flavor-impact chemicals in fermented pickles was compared using
solid-phase extraction (SPE), purge-and-trap (P&T) on Tenax, and SPME (5).
Determination of which chemicals were the most significant flavor-impact chemi-
cals was made by olfactometry experiments employing the detection frequency
method (for details, see Chapter 12) and by recombination studies (i.e., adding
suspect flavorants to an artificial salt brine solution to determine which potential
flavorant and/or combination of flavorants most closely matched the flavor of
the pickle brine).
The most potent odorants that define the typical characteristic brine aroma
were trans-4-hexenoic acid and cis-4-hexenoic acid. Confirmation of key impact
odorants in brine was made by recombination experiments. These highly odifer-
ous acids were detected in samples prepared by SPE and SPME (headspace using
a 75-m Carboxen/PDMS fiber) but not in samples prepared by P&T.
Chromatograms of the brine sample analyzed by P&T, SPE, and SPME
are illustrated in Fig. 4, with peak identities appearing in Table 4. These results
dramatically illustrate how the selection of the sample preparation technique in-
fluences the amounts and types of chemicals extracted and detected by an analyti-
cal procedure. P&T was able to detect a greater number of volatiles in the brine
than either SPE or SPME. Most of the volatiles that were detected by P&T and
not by SPE or SPME were early-eluting fusel oil components. Preliminary olfac-
tometry studies of volatiles extracted by P&T showed that the fusel oil fraction
contributed little significant olfactometry properties, and no single compound was
SPME Comparison Studies
TABLE 3 Comparison of the Principal Analytical Parameters for Pentanal and Hexanal Analyzed by DH GC/MS and SPME GC/MS
Repeatability of four
Analytical Detection limit replicates at 2 ng/mL Linear least squares
Compound Sample technique (ng/mL) (coefficient of variation, %) correlation coefficienta
Pentanal Skim DH 0.1 8.0 0.966

SPMEb 0.1 1.9 0.990
Hexanal Skim DH 0.3 21.1 0.910
SPME 0.5 7.1 0.995
Pentanal 2% Milk DH 0.3 7.6 0.996
SPME 0.3 2.1 0.999
Hexanal 2% Milk DH 0.8 8.3 0.982
SPME 0.8 4.9 0.993
a
For calibration curve of five standards ranging from 0.0 ng/mL to 30.0 ng/mL.
b
Carboxen/PDMS fiber.
215
216 Marsili
detected in P&T chromatograms that had odor characteristics that were similar
to the fermented brine samples. In general, P&T generated more chromatographic
peaks but few with significant, intense odor characteristics. In this case, more
peaks did not result in a better understanding of which chemicals were responsible
for the odor of fermented cucumbers.
FIGURE 4 GC/MS chromatograms of pickle volatiles obtained by three different sample

preparation techniques: (A) dynamic times headspace on Tenax with a 30-m 0.25-mm-
i.d. DB-5 column, film thickness 1 m; (B) solid-phase extraction (with C18 cartridges)
with a 30-m 0.25-mm-i.d. FFAP column, film thickness 0.25 m; and (C) headspace
SPME (75-m Carboxen/PDMS fiber) using same analytical column as in the dynamic
headspace method. Peak identities appear in Table 4.
Because P&T was apparently not extracting the impact odor chemical(s),
additional sample preparation techniques were investigated. SPE (with C18 car-
tridges) extracted fewer compounds and did a poor job extracting fusel oil compo-
nents. However, olfactometry experiments showed that SPE extracted more com-
pounds with strong aromas than P&T, and the odor characteristics of two
chemicals (cis- and trans-4-hexenoic acids) were observed to strongly match the
odor of brine samples. One problem observed with SPE was that degradation of
nonvolatile brine components (e.g., perhaps chlorophyll and other plant pig-
ments), which were not cleaned up by the SPE method, tended to elute as broad
peaks at the end of the chromatographic runs. This problem intensified as more
samples were analyzed by SPE. These dirty injections eventually resulted in
column fouling.
Initially, two types of SPME fibers were investigated75-m Carboxen/
PDMS and 70-m Carbowax/DVB Stable Flex. Both fibers were selected
for study because previously published work (7) indicated their appropriateness
for extracting organic acidscompounds that SPE studies had shown to be im-
TABLE 4 Peak Identification for Chemicals in Pickle Brine Analyzed by Various GC-MS Sample Preparation Techniques
Peak Analytical Peak Analytical
218
no.a Compound techniqueb no. Compound technique
1 Acetone P&T, SPME 27 Geranyl acetone P&T
2 Isopropanol P&T 28 5-methyl-2-furancarboxaldehyde (T) SPE
3 Ethyl acetate P&T, SPME, Z 29 Butyric acid SPE, SPME
4 Isobutyl alcohol P&T 30 Phenylacetaldehyde SPE
5 n-butyl alcohol P&T 31 Hexanoic acid SPE
6 Acetic acid P&T, SPME, Z 32 3-(methylthio)-1-propanol (T) SPE
7 2-pentanol P&T, Z 33 Acetamide (T) SPE
8 Isoamyl alcohol All 34 Benzyl alcohol SPE
9 1-pentene P&T 35 Phenyl ethyl alcohol SPE, SPME
10 1-pentanol P&T 36 trans-4-hexenoic acid SPE, SPME
11 Ethyl butyrate P&T, Z 37 cis-4-hexenoic acid SPE, SPME
12 2,3-butanediol P&T, SPME, SPE 38 Phenol SPE
13 1-hexanol P&T, Z 39 trans-2,4-hexadienoic acid SPE, SPME
14 Dihydro-4,5-dimethyl-2[3H]-furanone (Tc) P&T 40 Dimethyl sulfide SPME
41 Hexanal SPME, Z
15 Benzaldehyde P&T, SPE, Z 42 2-methyl-1-pentene SPME
16 Octanal P&T, Z 43 2-heptanol SPME, Z
17 Dichlorobenzene P&T 44 5-hepten-2-one (T) SPME
18 2-ethyl-1-hexanol P&T, SPE, Z 45 2,4-hexadienal SPME
19 Decanal P&T, Z 46 Nonanal SPME, Z
20 Linalool oxide (T) P&T 47 cis-2,4-hexadienoic acid SPME
21 Linalool P&T, SPME, Z 48 2,6-nonadienal SPME, Z
22 Undecyl aldehyde P&T 49 2-dodecen-1-al (T) SPME
23 -terpineol P&T, SPME B Artifact (not from sample)
24 Dodecyl aldehyde P&T U Unknown
25 Isothiocyanato cyclohexane (T) P&T S SPE eluting solvent (methanol) SPE
26 Tetradecanal P&T I.S. Int. Std. (4-methyl-2-pentanone) SPE
Marsili
a
Peak numbers apply to chromatograms in Fig. 4A, 4B, and 4C.
b
Analytical technique used to analyze brine sample. P&T purge-and-trap; SPE solid-phase extraction; SPME solid-phase microextraction; Z refers
to previously published results (6) based on purge-and-trap GC/MS (HP-5 capillary column).
c
Tentative identification based on mass spectrometry.
portant odorants in the brine. The two fibers performed about equally well, with
both extracting detectable levels of organic acids, even when present at low ppb
levels. The Carboxen/PDMS fiber demonstrated superior sensitivity for a few of
the more volatile compounds compared to the Carbowax/DVB fiber, and it ex-
tracted significantly more compounds than SPE. Therefore, SPME using the
Carboxen/PDMS fiber was ultimately selected.
Of the significant odor-active compounds observed in brine samples, only
cis- and trans-4-hexenoic acids and phenyl ethyl alcohol (or phenyl acetaldehyde,
an oxidation product of phenyl ethyl alcohol) were present in all brine samples
tested. (Note: Phenyl acetaldehyde has a floral/lilac/hyacinth aroma.) Therefore,
these highly odiferous chemicals are likely to be key impact-odor components of
fermented cucumbers. This was confirmed by detection frequency olfactometry
experiments and by recombination studies.
3. Conclusions
The milk and pickle studies comparing analytical results obtained by SPME with
other traditional sample preparation techniques illustrate the following:
1. SPME can provide quantitative and qualitative results as good as or
better than traditional sample preparation techniques.
2. PDMS is not always the fiber of choice.
3. Detecting more chromatographic peaks doesnt make one sample prep-
aration method better than another; it is more important to detect the
odor-active compounds, which may include high-boiling-point chemi-
cals, at their odor detection threshold.
4. SPME can be coupled with olfactometry techniques to determine the
primary odor-active compounds in a food system.
Some flavor chemists criticize SPME for not extracting a truly representa-
tive sample of the headspace gas above a food sample. These critics point out
that static headspace and possibly solvent extraction techniques do a better job
at extracting a profile of volatiles that is consistent with what people actually
detect when a food product is smelled. While it is true that for some applications
SPME demonstrates a bias for extracting higher boiling point compounds over
lower boiling point compounds, the higher boiling polar components are some-
times the odor-impact compounds of interest, and techniques such as static head-
space and solvent extraction are not sensitive enough to detect them. P&T tech-
niques in which the aroma chemicals are concentrated on an adsorbant (e.g.,
Tenax) are also biased in that stripping the liquid sample or flushing the sample
surface with a gas causes the most volatile components to be enriched compared
to the less volatile components, and the composition will not be representative
of the gas phase at equilibrium as it is perceived by the nose. Every sample
220 Marsili
preparation technique has its limitations and biases. However, if these limitations
and biases are understood, they can be used as an advantage rather than a liability.
In this pickle study, for example, the selectivity/bias of SPME is a benefit rather
than a detriment in performing flavor research studies.
III. THE IMPORTANCE OF SPME FIBER SELECTION
Roberts et al. (8) evaluated different types of SPME fibers for the analysis of
coffee and various flavored solutions. Experiments comparing different fibers
showed that PDMS/DVB had the highest overall sensitivity. Carboxen/PDMS
was the most sensitive to small molecules and acids.
Several factors, including the need to be in the linear range and competition
effects between volatile compounds, can cause biases in the quantitative determi-
nation of compounds (9). Blank et al. has shown that these limitations can be
overcome by using isotope-labeled compounds as internal standards (10).
1. Coffee. Three fibers were compared for the analysis of coffee volatiles:
PDMS, Carboxen/PDMS, and PDMS/DVB. Chromatograms of coffee volatiles
obtained from each of the three different types of fibers appear in Fig. 5. The
PDMS fiber showed the lowest overall sensitivity. Carboxen/PDMS was the most
FIGURE 5 Comparison of different fibers for the analysis of brewed coffee volatiles.
Conditions: 1 mL in 16 mL vial, 10 minutes headspace adsorption at room temperature,
followed by GC-MS using a DB-Wax column. Peak identities: (1) 2-methylpropanal; (2)
2-and 3-methylbutanal; (3) 2,3-pentanedione; (4) guaiacol; (5) ethylguaiacol; (6) vinyl-
guaiacol; (7) acetaldehyde.
sensitive for small molecules such as 2-methylpropanal and acetaldehyde. How-

ever, some peak tailing was observed with the Carboxen/PDMS fiber. PDMS/
DVB provided the best sensitivity, especially for molecules such as guaiacol, 4-
ethylguaiacol, and 4-vinylguaiacol.
2. Organic acids in water. Roberts et al. also compared how three fibers
(Carboxen/PDMS, Carbowax/DVB, and Polyacrylate) extracted volatile carbox-
ylic acids from aqueous solutions. Figure 6 shows that Carbowax/PDMS pro-
vided the best overall sensitivity for these compounds. However, Carboxen/
PDMS was more sensitive for the smaller acids (i.e., acetic and propionic).
3. Polar compounds of low volatility. Finally, Roberts et al. looked at
how furaneol, sotolon, and vanillinpolar, low volatility compoundswere ex-
tracted from aqueous model solutions by five different types of fibers. Of the
fibers compared, PDMS/DVB was found to give the best sensitivity for these
compounds.
Knowledge of the linear dynamic concentration range of an SPME method
is important for accurate quantitative results. The high sensitivity of SPME fibers
to most compounds leads to a linear range that is usually less than 1 ppm. In
natural products, all compounds that are present may not be in the linear range.
FIGURE 6 Comparison of different fibers for the analysis of volatile carboxylic acids.
Conditions: aqueous solution (0.2 mL) in 2 mL vial, 30C, pH 2, saturated with NaCl,
and 60 minutes headspace extraction. Concentrations of acids: acetic, 11 ppm; propionic,
0.7 ppm; isobutyric, 0.7 ppm; butyric, 0.8 ppm; isovaleric, 0.6 ppm; valeric, 0.7 ppm.
222 Marsili
Above a certain concentration level, there is less and less increase in the amount
of analyte bound to the fiber as the concentration increases. In order to compen-
sate for this problem, three techniques may be implemented:
1. Diluting the sample matrix
2. Reducing the adsorption time
3. Calibrating with isotope labeled standards that behave identically to
the target compound
For example, to work in the linear range for coffee analysis, Roberts et
al. used a short adsorption time. They found, however, that some rather polar
compounds (e.g., 2,3-pentanedione, pyrazine, furfural, and 2-acetylfuran) showed
slight nonlinear behavior.
A. Competition Phenomena
Competition between analytes for absorption/adsorption sites on the SPME fiber
is important to realize and control. Examples of competitive analyte behaviors
that have been previously observed include ethanol replacing acetone and iso-
prene on PDMS/DVB (11); methyl isobutyl ketone replacing 2-propanol, 2-
methyl-2-propanol, and tetrahydrofuran with Nafion-coated fibers (12); and al-
kylpyridines replacing pyridine with CW/DVB and PDMS fibers (9).
When Roberts et al. studied how added 2-isobutyl-3-methoxypyrazine in-
fluenced the ability of PDMS fibers to adsorb guaiacol, dimethyltrisulfide, 3-
methylbutanal, and 2,3-diethyl-5-methyl pyrazine, they found that up to 1 ppm
added pyrazine caused little effect. However, higher levels resulted in significant
decreases of the other compounds. Compounds that have high fiber-to-air equilib-
rium constants (e.g., 2-isobutyl-2-methoxypyrazine) are highly adsorbed by the
fiber and can, therefore, reduce the adsorption of other compounds.
Competition phenomena occur when the concentration exceeds the upper
limit of the linear range. Short sampling times are an effective way to reduce
fiber overloading and resulting biases, especially when both compounds of low
and high affinity for the fiber are analyzed.
Competition effects can adversely impact quantitation accuracy when two
samples are being compared for amounts of target volatiles. For example, if one
of the samples has a large concentration of a compound with a high fiber-to-air
equilibrium constant that the other does not, adsorption amounts of the target
volatiles would be affected. In this case, the two samples could not be accurately
compared unless isotopic standards were used for quantitation.
Roberts et al. and others found that when using shorter SPME extraction
times (1 min), the chromatographic profiles of conventional static headspace and
SPME headspace analysis were more similar than when 60-min SPME headspace
extraction times were used. However, the less volatile compounds are still de-
tected better with SPME, and the highly volatiles compounds are still detected
better with static headspace. Although results with shorter extraction times com-
pare more favorably to conventional static headspace results, short sampling
times with SPME have reduced sensitivity compared to longer extraction times.
If sensitivity is sufficient for a given analyte, shorter SPME sampling times are
usually preferred for quantitation, since there is a higher likelihood that there is
a linear relationship between the adsorbed analyte amount and the initial sample
concentration.
B. Fiber Selection Based on Analyte Characteristics

Using aqueous dilutions of a variety of standard organic compounds, R.E. Shirey
systematically studied SPME optimization of extraction conditions (including
fiber types) for low-molecular-weight analytes (13) and high-molecular-weight
volatiles and semivolatiles (14).
1. Low-Molecular-Weight Analytes
Shirey extracted a group of 13 volatile analytes with molecular weights under
90 amu using identical conditions and six different SPME fibers. Eleven classes
of organic compounds were represented in the group of volatiles studied. For the
extraction of small, low-molecular-weight molecules, the effects of SPME fiber
porosity far exceeded the effects of SPME fiber polarity and film thickness. The
micropores of the Carboxen-PDMS make it an ideal fiber for extracting these
analytes. The fiber is nonselective and extracted all of the analytes at magnitudes
better than the other SPME fibers. The only exception to this conclusion was for
isopropylamine. The ability of the PDMS-DVB coating to extract small amines
made the dual-layered DVB-Carboxen-PDMS the preferred fiber for isopropyl-
amine.
The effect of fiber polarity on small amines was minimal. The polar fibers
did not extract more of the polar analytes than the nonpolar analytes. However,
the polar fibers extracted much less of the nonpolar analytes than the nonpolar
fibers. The reduction in the extraction on nonpolar analytes by the polar fibers
provides some selectivity for polar analytes.
Basic compounds were best extracted from a solution at high pH levels,
and acidic compounds were best extracted from solutions at low pH levels. Sur-
prisingly, some analytes that are relatively neutral were extracted best in acidic
or basic solutions. It was observed that two chemicals in the stock mixture, nitro-
propane and methylacetate, were not stable in solutions at pH 11; this emphasizes
the importance of determining the effect of pH on the stability of analytes. In-
creasing ionic strength with 25% NaCl compared with deionized water improved
the recovery of all of the analytes, with polar analytes being more greatly affected
than nonpolar analytes.
224 Marsili
Heated headspace and immersion techniques proved best for extracting po-
lar analytes, and ambient headspace and immersion were the best techniques for
extracting nonpolar analytes. For an adsorbent-type fiber such as Carboxen, the
type of extraction used did not greatly affect the amount of analyte extracted.
For an absorbent-type fiber, the type of extraction used was more critical. Heated
headspace was best for extracting polar analytes, and either direct immersion or
ambient headspace were best for extracting nonpolar analytes. There are excep-
tions to these conclusions, but they serve as general guidelines for SPME extrac-
tion of low-molecular-weight compounds.
2. High-Molecular-Weight Volatiles and Semivolatiles

In another study, Shirey selected a group of 15 large-molecular-weight analytes
(MW 92-499 amu) and extracted them with nine different SPME fibers. Thirteen
classes of organic compounds were represented in this group of analytes.
Based on studies with the stock standard solution of 15 analytes, Shirey
recommended that two things be considered prior to the selection of an SPME
fiber:
1. The polarity and functionality of the analyte
2. The molecular weight and shape of the analyte
For semivolatile analytes, polarity with respect to fiber type was critical.
The more polar analytes were best extracted by polar fibers such as polyacrylate
and CW-DVB. The less polar analytes were extracted by both polar and nonpolar
fibers and, in some cases, were extracted better with the polyacrylate fiber. The
functional group that increases polarity was also important. Amines were ex-
tracted well with the PDMS-DVB fiber, whereas CW-DVB and polyacrylate were
better for other polar functional groups.
The size and shape of the analytes were also important. Smaller analytes
(200 amu) were more efficiently extracted by adsorbent-type fibers. The pores
offered the advantage of better retention of analytes compared with absorbent-
type fibers. Larger analytes (200 amu) or highly planar analytes such as PAH
were better extracted by absorbent-type fibers than adsorbent fibers. Layering the
DVB over Carboxen expanded the molecular weight range for improved recovery
when compared with the Carboxen fiber. However, large PAHs still were not
efficiently released during desorption. The bare fused-silica fiber was capable of
extracting nonpolar analytes but not reproducibly, and it had limited capacity.
When extraction techniques were compared, immersion of the fiber in the
water sample was superior to heated headspace sampling. However, the nonpolar
analytes (even those with high molecular weights and boiling points) could be
extracted using heated headspace. The recovery using heated headspace was 3
85% of the recovery obtained by immersion. The recovery of the polar semivola-
TABLE 5 SPME Fiber Selection Guide
Analyte class Fiber type Linear range
Acids (C2C8) Carboxen-PDMS 10 ppb1 ppm

Acids (C2C15) CW-DVB 50 ppb50 ppm
Alcohols (C1C8) Carboxen-PDMS 10 ppb1 ppm
Alcohols (C1C18) CW-DVB 50 ppb75 ppm
Polyacrylate 100 ppb100 ppm
Aldehydes (C2C8) Carboxen-PDMS 1 ppb500 ppb
Aldehydes (C3C14) 100-m PDMS 50 ppb50 ppm
Amines PDMS-DVB 50 ppb50 ppm
Amphetamines 100-m PDMS 100 ppb100 ppm
PDMS-DVB 50 ppb50 ppm
Aromatic amines PDMS-DVB 5 ppb1 ppm
Barbiturates PDMS-DVB 500 ppb100 ppm
Benzidines CW-DVB 5 ppb500 ppb
Benzodiazepines PDMS-DVB 100 ppb50 ppm
Esters (C3C15) 100-m PDMS 5 ppb10 ppm
Ethers (C4C12) Carboxen-PDMS 1 ppb500 ppb
Explosives (nitroaromatics) PDMS-DVB 1 ppb1 ppm
Hydrocarbons (C2C10) Carboxen-PDMS 10 ppb10 ppm
Hydrocarbons (C5C20) 100-m PDMS 500 ppt1 ppm
Hydrocarbons (C10C30) 30-m PDMS 100 ppt500 ppb
Hydrocarbons (C20C40) 7-m PDMS 5 ppb500 ppb
Ketones (C3C9) Carboxen-PDMS 5 ppb1 ppm
Ketones (C5C12) 100-m PDMS 5 ppb10 ppm
Nitrosamines PDMS-DVB 1 ppb200 ppb
PAHs 100-m PDMS 500 ppt1 ppm
30-m PDMS 100 ppt500 ppb
7-m PDMS 500 ppt500 ppb
PCBs 30-m PDMS 50 ppt500 ppb
Pesticides, chlorinated 100-m PDMS 50 ppt500 ppb
30-m PDMS 25 ppb500 ppb
Pesticides, nitrogen Polyacrylate 50 ppt500 ppb
Pesticides, phosphorus 100-m PDMS 100 ppt1 ppm
Polyacrylate 100 ppt500 ppb
Phenols Polyacrylate 5 ppb500 ppb
Surfactants CW-TPR 1 ppm100 ppm
Sulfur gases Carboxen-PDMS 10 ppb10 ppm
Terpenes 100-m PDMS 1 ppb10 ppm
VOCs Carboxen-PDMS 100 ppt500 ppb
100-m PDMS 20 ppb50 ppm
30-m PDMS 100 ppb50 ppm
Source: Courtesy of Dr. Robert Shirey, Supelco, Inc., Bellefonte, PA. From Solid-Phase Microextrac-
tion: A Practical Approach, 1999, S.A. Scheppers-Wercinski, ed., Marcel Dekker, Inc., New York.
226 Marsili
tile analytes using heated headspace was 0.650% that of the recovery by immer-
sion.
C. Fiber Selection Guide

One of the most important factors affecting analytical results with SPME is the
choice of fiber selected. At present, there are seven commercial fiber types avail-
able:
1. Poly(dimethylsiloxane) or PDMS (100-, 30-, and 7-m sizes)
2. Polyacrylate (85 m)
3. PDMS/divinylbenzene or DVB (65 m)
4. PDMS/Carboxen (75 m)
5. Carbowax/DVB (65 m)
6. Carbowax/template resin
7. Divinylbenzene/Carboxen/PDMS
Table 5 is a useful guide for selecting the optimum fiber for a given analyte
in specific concentration ranges (6).
IV. CONCLUSIONS
As numerous examples in this book illustrate, SPME is a potent sample prepara-

tion tool for the analysis of odor- and flavor-impact chemicals and is capable of
providing results equivalent to or better than conventional extraction techniques.
SPME is a deceptively simple technique. It is a fast and simple sample preparation
method, and sample chromatograms can be generated quickly. However, quanti-
tative accuracy with SPME is highly dependent on experimental conditions, sam-
ple matrix, analyte characteristics, the type of fiber, and calibration techniques.
SPME, more so than other extraction/concentration sample preparation tech-
niques, is an evolving technology because new fibers are constantly being intro-
duced.
SPME, like all extraction/concentration sample preparation techniques, has
strengths, weaknesses, and biases. Understanding what they are and how to con-
trol them allows users to maximize the effectiveness of SPME as a tool for the
analysis of flavors, fragrances, and odors.
REFERENCES

Food Chem. 42:1925 (1994).
2. F. Pelusio, T. Nilsson, L. Montanarella, R. Tilio, B. Larsen, S. Facchetti, and J.

Madsen, Headspace solid-phase microextraction analysis of volatile organic sulfur
compounds in black and white truffle aroma, J. Agric. Food Chem. 43:2138 (1995).
3. A. Krumbein and D. Ulrich, Comparison of Three Sample Preparation Techniques
for the Determination of Fresh Tomato Aroma Volatiles, in Flavour Science: Re-
cent Developments (A. J. Taylor and D. S. Mottram, eds.), The Royal Society of
Chemistry Information Services, 1996, p. 289.
4. R. T. Marsili, Comparison of solid-phase microextraction and dynamic headspace
method for the GC/MS analysis of light-induced lipid oxidation products in milk,
J. Chromatogr. Sci. 37:17 (1999).
5. R. T. Marsili and N. Miller, Determination of aroma-impact compounds in fermented
cucumbers by solid-phase microextractiongas chromatographymass spectrome-
tryolfactometry detection, J. Chromatogr. Sci. 38:307 (2000).
6. A. Zhou and R. F. McFeeters. Volatile compounds produced in cucumbers fermented
in low-salt conditions, J. Agric. Food Chem. 46:2117 (1998).
7. R. E. Shirey, SPME Fibers and Selection for Specific Applications, in Solid-
Phase Microextraction: A Practical Guide (S. S. Wercinski, ed.), Marcel Dekker,
New York, 1999, p. 83.
8. D. D. Roberts, P. Pollien, and C. Milo, Solid-phase microextraction method develop-
ment for headspace analysis of volatile flavor compounds, J. Agric. Food Chem. 48:
2430 (2000).
9. W. M. I. Coleman, A study of the behavior of Maillard reaction products analyzed
by solid-phase microextractiongas chromatographymass selective detection, J.
Chromatogr. Sci. 34:213 (1996).
10. I. Blank, C. Milo, J. Lin, and L. Fey, in Flavor Chemistry: 30 Years of Progress
(R. Teranishi, E. L. Wick, and I. Hornstein, eds.), Kluwer, Hingham, MA, 1999,
p. 63.
11. C. Grote and J. Pawliszyn, Solid-phase microextraction for the analysis of human
breath, Anal. Chem. 69:587 (1997).
12. T. Gorecki, P. Martos, and J. Pawliszyn, Strategies for the analysis of polar solvents
in liquid matrixes, Anal. Chem. 70:19 (1998).
13. R. E. Shirey, Optimization of extraction conditions for low-molecular-weight ana-
lytes using solid-phase microextraction, J. Chromatogr. Sci. 38:109 (2000).
14. R. E. Shirey, Optimization of extraction conditions and fiber selection for semivolat-
ile analytes using solid-phase microextraction, J. Chromatogr. Sci. 38:279 (2000).
8
Analysis of Volatile Compounds
in the Headspace of Rice Using
SPME/GC/MS
Casey C. Grimm and Elaine T. Champagne

U.S. Department of Agriculture, New Orleans, Louisiana
Kenichi Ohtsubo
Ministry of Agriculture, Forestry and Fisheries, Ibaraki-Ken, Japan
I. INTRODUCTION
Quality control of agricultural crops has traditionally been based on human sen-
sory perception. Although some instrumental methods exist, they are time con-
suming and expensive, often provide insufficient sensitivity, and generally are
impractical for large-scale analyses required in monitoring the quality control of
agricultural products. However, with recent technological advances, new efforts
have been directed toward applying modern analytical instrumentation to the as-
sessment of sensory quality. Currently, only a rudimentary understanding of the
relationship between chemical composition and sensory quality exists. This is
primarily the result of the presence of compounds that impact aroma when present
at concentrations approaching the parts-per-billion range. Not all of the these
compounds have yet been identified. Before automated instrumental methods can
be used to routinely access sensory quality, the compounds that contribute to
these aromas must be characterized.
To be perceived, compounds by nature must be somewhat volatile: they
need to traverse the mouth to reach and stimulate the olfactory bulb. Hence,
analysis of the volatile compounds present in the headspace over rice should
yield the desired information. Unfortunately, no one instrumental technique has
yet been developed that can simulate the response of the human nose. All instru-
229
230 Grimm et al.
mental analytical techniques have limitations and detect only a subset of the total
number of volatile compounds present in the headspace. Nevertheless, headspace
analysis yields a gas chromatographic profile that can be used for testing the
relationship between the relative concentration of sensory impact compounds to
overall sensory quality.
A. SPME
Standard methods of analysis employ schemes to capture volatile compounds,
concentrate them, separate them, and quantify them. Recent reviews of sample
preparation for volatiles analysis include foods in general (1), cereals (2), and
dairy products (3). Recent advances in analytical instrumentation and methodol-
ogy have approached the threshold of selectivity and sensitivity demonstrated by
the human nose. In this chapter, we describe new methodology employed to mea-
sure some of the key odorants in rice. The effective collection and analysis of
volatile compounds can now be accomplished using solid-phase microextraction
(SPME). A pictorial outline of the procedure is depicted in Fig. 1.
SPME has proven to be a simple, rapid, and sensitive method for collecting
the volatile compounds from the headspace of a sample (4,5). In theory, under
ideal conditions, the volatile compounds in a liquid sample partition between the
liquid and gas phase and between the gas phase and the SPME fiber. After suffi-
cient time, equilibria are established between the sample and the headspace and
the headspace and the SPME fiber. In liquid samples, the equilibria can be shifted
by adjusting the temperature, pH, using mechanical mixing, by the addition of
salts such as NaCl, or a combination of these treatments.
FIGURE 1 Block diagram of SPME analysis.

SPME/GC/MS and Volatile Compounds in Rice 231
The SPME fiber can be directly immersed into aqueous solutions or can
be deployed in the headspace above the sample. Direct immersion is not possible
when analyzing solid samples. Because the rice samples used in these experi-
ments were in solid form, this discussion will focus only on headspace analysis
by SPME.
In a sample consisting of a complex matrix such as food, the ideal condi-
tions are not readily achieved because the absolute concentration of a given com-
pound is not fixed. Compounds are continually formed and degraded through
myriad complex chemical reactions as a function of temperature, moisture, and
pressure. However, relative quantitative information can be obtained by treating
the samples under consistent conditions and measuring the resultant production/
release of the volatile compounds. A review employing SPME for the analysis
of foods has recently been published (6).
B. Rice
Although a large number of compounds have been observed in cooked rice
aroma, and over 100 have been identified, only a few have been found to have
an impact on the flavor of cooked rice (79). Tsugita concluded in his review
of cooked rice aroma that no single compound has been found that can be de-
scribed as cooked rice, and the aroma probably arises from a mixture of sev-
eral compounds (7). The majority of compounds present in the headspace of
cooked rice are lipid oxidation products (11,12). These compounds possess odor
threshold values in the high parts-per-million range and contribute minimally to
the aroma of cooked rice (10,13). However, with deterioration during storage,
the lipid oxidation products begin to exceed levels of sensory thresholds and
contribute an off-flavor commonly referred to as old or stale rice.
Yajima identified -pyrrolidone as a key odorant in Katorimai (scented
riceO. sative japonica) and noted the presence of indole (14). Buttery listed
64 volatile compounds known in rice and identified seven compounds with low
odor thresholds: octanal, nonanal, (E)-2-nonenal, decanal, (E)-2-decenal, (E,E)-
2,4-decadienal, and 2-acetyl-1-pyrroline (2-AP) (8). The latter compound, 2-AP,
is the major odorant contributor of scented or popcorn rice (15,16). Although
present in most rice at the 110 ppb level, in scented rice it can be found at
concentration levels in excess of 2 ppm (17). Widjaja, in a comparative study of
non-fragrant and fragrant rice, identified (E)-2-decenal, (E,E)-2,4-nonadienal,
and (E,E)-2,4-decadienal as having a waxy aroma (13). These three lipid oxi-
dation products are also found in glutinous or waxy rice and contribute to its
distinctive odor (18).
Comprehensive research programs have been initiated to gain a thorough
understanding of the key physical parametersappearance, texture, flavor, and
cooking propertiesthat constitute rice quality. This study was undertaken to
232 Grimm et al.
establish the relationships of the presence of compounds found in the headspace

with particular notes in the aroma of the cooked rice. A secondary goal of these
experiments was to establish a database consisting of the volatile compounds
observed from diverse rice cultivers grown under various conditions.
II. INSTRUMENTATION
This section describes the equipment used for the analysis by SPME. The me-
chanical application of SPME is analogous to liquid injection by hand. The
startup costs are small and a vast amount of information can be gained rather
quickly. Typically, for a laboratory already performing GC analysis, all that is
needed are SPME fibers and an SPME holder. A pack of fibers and a holder can
be obtained for several hundred dollars. No modifications to a GC are needed.
It is recommended that a reduced inlet liner (0.7 mm I.D.) be used in the injection
port, although SPME can be successfully performed with a standard inlet liner.
The injection of 1 l of liquid sample will rapidly expand to fill the 1 ml volume
of a typical inlet liner. Because there is no liquid solvent being injected with
SPME, there is no need for the relatively large inlet volume. The reduced volumes
in the injection port produce narrow injection bandwidths resulting in sharper
GC peaks.
Once large-scale sample throughput is required, automated systems are
needed for sampling and in some cases for preparation. Costs for this equipment
are on par with traditional autosampler systems, and the base costs run $15,000
to $20,000. Additional accessories such as Peltier cooling devices and needle
heaters can add an additional $5000. Multipurpose autosamplers are available
that can perform standard liquid injections, headspace sampling, and SPME.
A. Manual SPME
A comprehensive book on the theory and practice of SPME has been put together
by Pawlizyn (19). For developing methodology using SPME, initial experiments
should employ volatile standards in an aqueous solution in the 110 ppb range.
Even the presence of small amounts of solvent can readily affect the recovery
of standards. Samples should be temperature controlled with the use of a water
bath or other thermostated device. Sample vials should be cleaned and baked out
to remove contaminants absorbed from the packaging during shipping. Contami-
nants such as phthalic acid esters are readily observed using SPME. For a 2-ml
vial, with 1 ml of solvent, the presence of 1 ng of material results in a solution
at a concentration of 1 ppb. If a routine liquid aliquot of 1 l of that solution is
injected, only 1 picogram of material would effectively be placed on the GC
column. This amount of material is generally lost in the baseline of the GC trace.
With SPME, recoveries greater than 90% of some compounds can be collected
by the fiber and result in the placement of nanogram quantities of analyte on the
GC column. Hence, standards are routinely made at the ppb level for SPME in
contrast to ppm levels for liquid injections.
For optimal recovery, the aim is to minimize the headspace but to leave
sufficient room for the deployment of the SPME fiber. In our laboratory, we have
observed maximal recoveries of standards using a saturated solution of NaCl, a
thermostated water bath, and a magnetic stir bar. The magnetic stir bar is placed
inside the sample vial along with NaCl. The volume of the sample is adjusted
so that the fiber is deployed into the vortex created by the spinning magnet. If
employing a 12-ml sample vial, the use of a 2-cm length SPME fiber can be used
to enhance recovery.
B. Automated SPME
The Varian Corporation was the first company to produce an automated SPME
system. They adapted their liquid autosampler to run SPME. The initial system
contained an agitator, but no means of controlling the sample temperature. The
construction of a cover placed over the sample tray and use of the rising heat
from the GC system resulted in a thermostated region of approximately 35C.
Subsequently, a thermostated heated tray was introduced by Varian. Two sample
trays sizes were available, a tray for 2-ml vials holding 48 samples or a tray for
12-ml vials holding 12 samples. Shortcomings of the system were the agitation
of the SPME fiber in the headspace (resulting in a much weaker recovery of the
analyte) and the heating of the whole tray of samples while waiting for analysis
(which could result in the formation of different volatiles in certain samples, i.e.,
rice). The software operates on a Windows 3.1 platform and is not compatible
with Windows 95 or Windows NT.
Leap Technologies, which represents the Swiss CTC company in the
United States, added SPME capabilities to their combination headspace/liquid
autosampler. Samples are maintained at room temperature prior to sampling in
one of several trays. Tray sizes include a 32 12 ml tray that can be flipped
over to accommodate 21 2 ml samples. A separate 98 2 ml tray is also
available. A Peltier cooling device can be used in conjunction with the 12-ml
samples but does not work with the 98 2 ml trays. This allows samples to be
maintained at refrigerated temperatures until sampled. An external needle heater
is available for additional thermal cleaning of the fiber. However, initial models
had a problem with maintaining the set temperature, resulting in rapid deteriora-
tion of the SPME fibers. Special caps are required for sample vials because the
robot arm employs a magnet to pick up the vials and place them into a shaker
that can be heated. The SPME fiber is then deployed while the sample is being
shaken. Adsorption parameters can be selected and stored as a method for later
234 Grimm et al.
retrieval. At present no other SPME autosamplers are commercially available,

although the CTC instrument can be obtained from most GC vendors.
III. QUALITATIVE ANALYSIS
In most cases, SPME is readily amenable for use in qualitative analysis. However,
initial efforts to use SPME to measure rice volatiles proved less than satisfactory
(20). The limitation was that the initial stationary phases used as coatings for
SPME fibers were limited to temperatures slightly above ambient. Placing the
fiber in the effluent (dynamic SPME) of a sample purged with nitrogen improved
the recovery of volatile compounds. However, quantitation was problematic be-
cause the additional variables of purge flow rate, path, and time contributed to
increased variance.
The introduction of SPME fibers with the capability of operating at higher
temperatures has resulted in improved collection of volatile compounds from the
headspace of rice. Compounds not observed at room temperature can now be
desorbed, collected, and concentrated in a single step. Additionally, heating the
sample shifts the concentration equilibrium toward the gas phase, thus enhancing
sensitivity.
The total ion chromatogram of the analysis of a Thai jasmine rice is shown
in Fig. 2. The samples was desorbed for 5 minutes on a HP 5973 GC/MS system
(Agilent Technologies, Palo Alto, CA). The injector temperature was held con-
stant at 270C. The GC oven temperature was held for 1 min at 50C, then ramped
to 250C at 10C/min. A 30 m, 0.25 mm I.D., DB-5 capillary column with 1.0
m film was used with helium as the carrier gas under a constant flow of 40
cm/s. The total GC cycle time consisted of a 30-minute run and a 5-minute cool-
down period. Following the first GC/MS run, subsequent samples were prepared
ahead so that one sample was run every 45 minutes. The mass spectrometer was
operated in scan mode from m/z 50 to m/z 350.
A. Sample Preparation
Rice is typically harvested at approximately 20 to 25% moisture and dried to a
final moisture content of 1015% to inhibit germination and/or growth of micro-
organisms. At this stage, the rice is still in the hull and is referred to as rough
rice or paddy rice. Following drying, rice is dehulled to yield brown rice. In the
United States, rice is generally milled first and then stored. In Japan, rice is gener-
ally stored as brown rice and is milled only a few days prior to use. Removal
of the bran results in white rice, which contains only the starchy endosperm of
the original rice kernel. Rice can be analyzed as paddy, brown, milled rice, or
rice flour.
Research on rice volatiles is generally focused on white rice, although some

research has been conducted on brown rice and on paddy rice (7,21). The volatile
profile from the headspace of rice can be readily changed by the addition of water
and cooking the rice. Varying amounts of water in cooked rice results in signifi-
cant changes in the GC profile (22). In the case of 2-acetyl-1-pyrroline (2-AP),
the addition of a small amount of water results in an increase in the recovery of
2-AP. Figure 3 shows the recoveries of rice with 0, 100, 250, 500, and 750 l
of water added to 0.75 g of rice. Peak areas were integrated using the base peak
at m/z 83. The addition of 100 l of water increases the recovery of 2-AP fivefold.
Further addition of water results in a decrease in the amount of 2-AP observed.
The addition of water tends to increase the recovery of palmitic and oleic acid.
For targeted analysis such as described for 2-AP, the addition of a small amount
of water is advantageous, whereas for other compounds the addition of water
may suppress their recovery. The addition of water can further complicate analy-
sis because it can induce enzymatic action and accelerate metabolic and catabolic
functions. If the samples are prepared with water and stored over several days,
growth of microorganisms can be observed in some samples and the GC profiles
can be compromised.
The volatile profile changes dramatically when comparing white rice and
rice flour. In the rice flour, a large increase is observed in lipid oxidation products
such as hexanal. An increase is also observed in the abundance of 2-AP, even
though no water is added. However, the amount of 2-AP released from rice flour
is similar to that from milled rice kernels when 100 l of water is added. Addition
of water to the rice flour results in no significant increase in 2-AP while reducing
the amount of hexanal recovered.
B. SPME Optimization
Following preliminary SPME analysis, it is normal practice to optimize the
method for reproducibility, sensitivity, and ease of use. For rice, this consisted
of varying the parameters of preheating time, sample/vial size, adsorption time,
and temperature, in addition to the different sample preparations for rice, as dis-
cussed in the preceding section. Due to the higher sampling temperatures, only the
Carboxen/DVB/PDMS fiber was considered. When performing manual SPME, it
is possible to maximize the sample/headspace ratio by employing a large sample
container and minimizing the headspace to enhance recovery.
With an autosampler the available vial size is dictated by what vials will
fit in the trays of the autosampler. For our systems this included a 2-ml vial and
a 12-ml vial. When possible, the sample volume should be adjusted to minimize
the headspace, leaving just enough room to deploy the SPME fiber. For the 12-
ml vials, this results in a sample size of 5 to 6 g of rice occupying approximately
two-thirds of the volume. For the 2-ml vials, the sample is limited to 0.75 g of
236 Grimm et al.
rice occupying approximately 50% of the volume. Ths smaller vial utilizing only
0.75 g of rice gave results similar to the larger sample and was selected for all
future experiments. Methodology developed using the smaller amount also allows
for the repeated analysis of experimental rice from breeders where only limited
amounts are available. Rice kernels weigh between 0.02 and 0.03 g each and
thus, on average, there are 25 kernels per sample. This is a sufficient number of
kernels for a representative analysis, assuming that there was a sufficient mixing
of the rice kernels prior to removing the aliquot.
Adsorption times of 15 to 30 minutes are typical for SPME analysis. Com-
pounds possessing greater volatility will take less time to equilibrate than those
of lesser volatility. For rice kernels an adsorption time of 15 minutes gave satis-
factory results, but a preheating time of 20 minutes was required.
C. Observed Compounds
The chromatographic traces from the analysis of the headspace of rice can be
deceiving. The stark differences between two samples may initially appear to
provide an easy method for varietal, location, and growing differences. However,
on close examination, the differences usually are a result of lipid oxidation prod-
ucts and contaminants, which occur during post-harvest handling and storage
processes. True differences that can be used to distinguish between the aromas
of rice are quite subtle.
1. Contaminants
As with any trace level analysis, the GC profile will contain certain compounds
that are introduced by the method. In the analysis of rice using a carboxen/DVB/
PDMS fiber, certain compounds are routinely seen that do not originate from the
sample. The molecular weights (MW), molecular formulas (MF), chemical ab-
stract number (CAS), prominent ions from the mass spectrum along with relative
abundance (in parenthesis), and probable source of origination are given in Table
1. A series of siloxanes are regularly observed in the GC trace. In Fig. 2, they
are marked with an asterisk. Possible sources of siloxane contamination are the
PDMS in SPME fiber, the silinized inlet liner, the injector septum, and the capil-
lary column. The introduction of moisture into the injection port by the SPME
fiber would readily result in contribution of these compounds from all three
sources. With the possible exception of the first siloxane, they are cyclic in nature
and have a general formula of C2nH6nSinOn with 3n8 observe. It is possible
that larger cyclosiloxanes could be observed at higher column temperatures.
However, baking the caps and vial seals greatly reduces the level of these cyclic
contaminants. The first siloxane has a characteristic ion at m/z 77. A tentative
explanation for this peak is that it results from dihydroxy dimethyl silane at MS
SPME/GC/MS and Volatile Compounds in Rice
TABLE 1 Common Contaminants Observed During the Analysis of Rice Using a Carboxen/DVB/PDMS Fiber
Compound RT CAS # MW MF Mass (Relative Intensity) Origin
Unknown (siloxane) 704 92 77(100), 45(11) PDMS

Cyclotrisiloxane, hexamethyl- 780 541-05-9 222 C6H18O3Si3 207(100), 133(7), 96(10) PDMS
Tetrachloroethylene 806 127-18-4 164 C2Cl4 166(100), 129(60), 94(25) Unknown
oxime, methoxy-phenyl- 897 151 C8H9NO2 151(46), 131(100), 77(13) Fiber-Glue?
Cyclotetrasiloxane, octamethyl- 979 556-67-2 296 C8H24O4Si4 281(100), 164(10), 249(9) PDMS
Dichlorobenzene 1016 104-46-7 146 C6H4Cl2 148(65), 146(100), 111(30) Unknown
Cyclopentasiloxane, decamethyl- 1139 541-02-6 370 C10H30O5Si5 355(100), 267(75), 73(75) PDMS
BHT 1510 128-37-0 220 C15H24O 220(20), 205(100) Antioxidant
Ionol 2 1560 4130-42-1 234 C16H26O 234(22), 219(100) Antioxidant
Diethyl phthalate 1593 84-66-02 222 C12H14O4 222(4), 177(26), 149(100) Plasticizer
Cyclohexasiloxane, dodecamethyl- 1621 540-97-6 444 C12H36O6Si6 429(25), 341(40), 73(100) PDMS
Unknown 1680 100(100) Vespel Ferrule
Cycloheptasiloxane, tetradecamethyl- 1772 107-50-6 518 C14H42O7Si7 503(40), 281(80), 73(100) PDMS
Dibutyl phthalate 1867 84-74-2 278 C16H22O4 278(2), 205(7), 149(100) Plasticizer
Unknown 1894 100(100) Vespel Ferrule
Cyclooctasiloxane, hexadecamethyl- 1913 556-68-3 592 C16H48O8Si8 577(15), 355(100), 73(90) PDMS
Bis[2-ethylhexyl) phthalate 2753 117-81-7 390 C24H38O4 279(25), 167(35), 149(100) Plasticizer
237
238 Grimm et al.
FIGURE 2 Total ion chromatogram of the volatile compounds observed in the headspace
of a Thai jasmine rice sample.
of 92, which upon ionization readily loses a CH3 radical to give the fragment
ion of m/z 77.
The compound at a retention index (RI) of 897 is tentatively identified as
a methoxy-phenyl oxime and possesses a base peak at m/z 133 and a molecular
ion of m/z 151. This compound is believed to originate from the glue that is used
to attach the fiber to the syringe plunger. A series of compounds are observed
that have a base peak of m/z 100. These compounds may be associated with
the action of moisture with the Vespel ferrules. Other contaminants include the
ubiquitous phthalic acid esters and antioxidants, butylated hydroxy toluene and
butylated hydroxy anisole. Sources of these compounds are the packaging used
for vials.
2. Endogenous Compounds in White Rice

Table 2 provides a listing of the compounds we have observed in a variety of
rice samples from throughout the world. Of all the compounds listed in Table 2,
only the variation of the concentration in the amount of 2-acetyl-1-pyrroline has
been positively shown to have a major impact on the aroma of rice. The origin
TABLE 2 Compounds Observed in the Headspace of Rice
Compound R.I. MW MF CAS #

Ethanol 250 46 C2H6O 64-17-5
Acetone 259 58 C3H6O 67-64-1
Dimethyl sulfide 308 62 C2H6S 75-18-3
Hexane 600 86 C6H14 285-58-5
Butanal 602 72 C4H8O 123-72-8
Acetic acid 622 60 C2H 4O 2 64-19-7
3-Methyl-butanal 652 86 C5H10O 590-86-3
2-Methyl-butanal 660 86 C5H10O 96-17-3
Heptane 700 100 C7H16 142-82-5
Pentanal 701 86 C5H10O 110-62-3
Methyl butanoate 710 102 C5H10O2 623-42-7
3-Methyl-butanol 730 88 C5H12O 123-51-3
2-Methyl-butanol 730 88 C5H12O 137-32-6
Dimethyl disulfide 741 96 C2 H 6 S2 624-92-0
Pyridine 763 79 C5H5N 110-86-1
3-Hexanone 744 100 C6H12O 589-38-8
1-Pentanol 761 88 C5H12O 71-41-0
Toluene 764 92 C7H8 108-88-3
2,4-Pentandione 778 100 C5H 8O 2 123-54-6
2,3-Butandiol 788 90 C4H10O2 513-85-9
1,3-Butandiol 791 90 C4H10O2 107-88-0
Octane 800 114 C8H18 111-65-9
Hexanal 800 100 C6H12O 66-25-1
Methyl pentanoate 821 116 C6H12O2 624-24-8
Ethenyl cycohexane 832 108 C8H12 100-40-3
Butanoic acid 638 88 C4H 8O 2 107-92-6
(E)-2-hexenal 850 98 C6H10O 6728-26-3
(Z)-3-hexenal 853 98 C6H10O 6789-80-6
Ethyl benzene 858 106 C8H10 100-41-4
1-Hexanol 865 102 C6H14O 111-27-3
1,4-Dimethyl benzene 868 106 C8H10 106-42-3
Trimethylheptane 874 142 C10H22 14720-74-2
Pentanoic acid 879 102 C5H10O2 109-52-4
2,2,4-Trimethylheptane 883 142 C10H22 14720-74-2
2-Heptanone 888 114 C7H14O 110-43-0
2-Butylfuran 890 124 C8H12O 4466-24-4
Styrene 892 104 C 8H 8 100-42-5
1,2-dimethylbenzene 893 106 C8H10 94-47-6
Nonane 900 128 C9H20 111-84-2
Heptanal 902 114 C7H14O 111-71-7
2-Butoxyethanol 905 118 C6H14O2 111-76-2
2,3,6-Trimethylheptane 913 142 C10H22 4032-93-3
240 Grimm et al.
TABLE 2 Continued

2-Acetyl-1-pyrroline 920 111 C6H9NO 85213-22-05
Methyl hexanoate 922 130 C7H14O2 106-70-7
alpha-pinene 932 136 C10H16 80-56-8
1-Butoxy-2-propanol 938 132 C7H16O2 5131-66-8
(E)-2-Heptanal 956 112 C7H12O 57266-86-1
1-Ethyl, 4-methylbenzene 959 120 C9H12 611-14-3
Benzaldehyde 962 106 C 7H 6O 100-52-7
1-Heptanol 969 116 C7H16O 111-70-6
Dimethyl trisulfide 970 126 C 2 H 6 S3 3658-80-8
Hexanoic acid 983 116 C6H12O2 142-62-1
6-Methyl-5-hepten-2-one 983 126 C8H14O 110-93-0
2-Pentylfuran 990 138 C9H14O 3777-69-3
2,3,6-Trimethylpyridine 989 121 C8H11N 1462-84-6
1,2,4-Trimethylbenzene 994 120 C9H12 95-63-6
Ethyl hexanoate 998 144 C8H16O2 123-66-0
Decane 1000 142 C10H22 124-18-5
Octanal 1004 128 C8H16O 124-13-0
2-Ethyl-1-hexanol 1029 130 C8H18O 104-76-7
Limonene 1030 136 C10H16 138-86-3
Indane 1035 118 C9H10 496-11-7
Benzyl alcohol 1036 108 C 7H 8O 100-51-6
(E)-3-Octene-2-one 1038 126 C8H14O 18402-82-9
Benzeneacetaldehyde 1045 120 C 8H 8O 122-78-1
5-Ethyldihydro-2(3H)-furanone 1049 114 C6H10O2 695-06-7
2-Octenal 1058 126 C8H18O 111-87-5
1-Octanol 1071 130 C8H18O 111-87-5
2-Methyl-1-propenyl benzene 1082 132 C10H12 768-49-0
1-Methyl,2-(1-methylethyl)-benzene 1084 134 C10H14 527-84-4
Undecane 1100 156 C11H24 1120-21-4
Nonanal 1106 142 C9H18O 124-19-6
1,2,3,4-Tetramethylbenzene 1123 134 C10H14 488-23-3
Methyl octanoate 1128 158 C9H18O2 111-11-5
5-Ethyl-6-methyl-(E)3-hepten-2-one 1147 154 C10H18O 57283-79-1
3-Methyl-2-heptyl acetate 1151 172 C10H20O2 72218-58-7
5-Propyldihydro-2(3H)-furanone 1156 128 C7H12O2 105-21-5
Beta-terpineol 1157 154 C10H18O 138-87-4
E-2-nonenal 1166 142 C9H16O 18829-56-6
Nonanol 1176 144 C9H20O 143-08-8
Naphthalene 1190 128 C10H8 91-20-3
Ethyl octanoate 1196 172 C10H20O2 106-32-1
Alpha-terpineol 1198 154 C10H18O 12/2/38
Dodecane 1200 170 C12H26 112-40-3
TABLE 2 Continued

Gamma-terpineol 1203 154 C10H18O 586-81-2
Decanal 1207 142 C10H22 112-31-2
E,E-2,4-nonadienal 1218 138 C9H14O 5910-87-2
Benzothiazole 1234 135 C7H5NS 85-16-9
2-Hexyl-1-octanol 1254 214 C14H30O 19780-79-1
5-Butyldihydro-2(3H)-furanone 1259 142 C8H14O2 104-50-7
E-2-Decenal 1267 154 C10H18O 3913-81-3
1-Tridecene 1293 182 C13H26 2437-56-1
3-(t-Butyl)-phenol 1295 150 C10H14O 585-34-2
Ethyl nonanoate 1296 188 C11H22O2 123-29-5
2-Undecanone 1300 170 C11H22O 112-12-9
Indole 1300 117 C 8H 7N 120-72-9
Tridecane 1300 184 C13H28 629-50-5
Undecanal 1310 170 C11H22O 112-44-7
Methyl decanoate 1325 186 C11H22O2 110-42-9
5-Pentyldihydro-2(3H)-furanone 1368 156 C9H16O 104-61-0
2-Butyl-2-octenal 1375 182 C12H22O 13019-16-4
Tetradec-1-ene 1392 200 C14H28 1120-36-1
Ethyldecanoate 1394 200 C12H24O2 110-38-3
Tetradecane 1400 198 C14H30 629-59-4
Isolongifolene 1408 204 C15H24 1135-66-6
trans-caryophylene 1433 204 C15H24 87-44-5
Bis-(1-methylethyl)-hexadecanoate 1448 230 C12H22O2 6938-94-9
(E) 6,10-dimethyl-5,9-undecadien-2-one 1450 194 C13H22O 3879-26-3
2,6-Bis(t-butyl)-2,5-cyclohexadien-1,4-dione 1471 220 C14H20O2 719-22-2
1-Hexadecene 1492 224 C16H32 629-73-2
Pentadecane 1500 212 C15H32 629-62-9
BHT 1510 220 C15H24O 128-37-0
Methyl dodecanoate 1523 214 C13H26O2 111-82-0
1-S, cis-calamene 1534 202 C15H22 483-77-2
Ionol 2 1561 234 C16H26O 4130-42-1
Ethyl dodecanoate 1592 228 C14H28O2 106-33-2
Hexadecane 1600 226 C16H34 544-76-3
2,6-bis-(t-butyl)-2,5-cyclohexadien-1-one 1637 232 C16H24O 6378-27-8
2,5,10-Trimethylpentadecane 1646 254 C18H38 3892-00-0
Heptadecane 1700 240 C17H36 629-78-7
2,5,10,14-Tetramethylpentadecane 1703 268 C19H40 1921-70-6
Methyl tetradecanoate 1724 242 C15H30O2 124-10-7
Indene, 2,3-dihydro-1,1,3-trimethyl-3-
phenyl 1731 236 C18H20 3910-35-8
2,6-Diisopropylnaphthalene 1744 212 C16H2O 24157-81-1
2-Methyl-2,4-diphenylpentane 1762 238 C18H22 31516-55-9
242 Grimm et al.
TABLE 2 Continued

Ethyl tetradecanoate 1793 256 C16H32O2 124-06-1
Octadecane 1800 254 C18H38 593-45-3
2,6,10,14-Tetramethylhexadecane 1807 282 C20H42 638-36-8
6,10,14-Trimethyl-2-pentadecanone 1844 268 C18H36O 502-69-2
Nonadecane 1900 268 C19H40 629-92-5
Methyl hexadecanoate 1925 258 C17H34O2 112-39-0
Ethyl hexadecanoate 1993 284 C18H36O2 628-97-7
Hexadecanoic acid 1995 256 C16H32O2 57-10-3
Eicosane 2000 282 C20H42 112-95-8
Methyl linolate 2051 294 C19H34O2 2566-97-4
Methyl oleate 2052 296 C19H36O2 112-62-9
Ethyl linoleate 2081 308 C20H36O2 544-35-4
Ethyl oleate 2086 310 C20H38O2 111-62-6
of some of these compounds has yet to be ascertained and are therefore suspect as
to whether they are endogenous or result from post-harvest handling and storage
processes. Some compounds such as styrene and butylated hydroxy toluene
(BHT) are adsorbed by the rice from their storage containers. The levels of lipid
oxidation products will vary depending on the robustness of the variety and the
storage time and temperatures. The relative amounts of low-boiling compounds
can vary depending on sample preparation and the amount of time allowed for
the sample to remain in a static headspace environment prior to sampling.
By far, the GC profile is dominated by the presence of lipid oxidation prod-
ucts. These compounds include most of the aldehydes, alcohols, and acids. They
originate from the breakdown of oleic and linoleic acid. These compounds in-
clude the straight chain aldehydes from pentanal to dodecanal, the alcohols from
pentanol to dodecanol, and in very rancid samples, the short chain fatty acids
from acetic to decanoic acid. There is a series of unsaturated aldehydes ranging
from (E)-2-heptenal to (E)-2-decenal. These compounds may originate as the cis
isomer in the 3 and 4 position but quickly isomerize to the more favorable trans
position.
3. Brown Rice
Brown rice contains a greater amount of lipid than white rice. The bran, present
on the brown rice but milled away from the white rice, is rich in lipids relative
to the starchy endosperm of the milled rice. Figure 3 compares selected volatile
compounds found in the headspace of brown rice and milled white rice. The
FIGURE 3 Relative intensities of the average of five milled rice samples and five brown
rice samples.
values are averaged from five different varieties. Lipid oxidation products such
as hexanal, (E)-2-heptenal, 2-pentylfuran, and hexanoic acid are found in greater
concentration in the milled rice than in fresh brown rice, even though milled rice
contains relatively little lipid. The milling process damages the tissue, and these
compounds are readily formed. The 2-AP constitutes only a small fraction of the
total headspace, yet this amount is sufficient to dominate the aroma of the cooked
rice. There appears to be only a slight difference between the levels of 2-AP
recovered from milled and brown rice samples. This is in agreement with a previ-
ous study that showed little difference in the 2-AP content of brown and milled
rice (23). Nonanal and naphthalene show slightly greater concentrations in brown
rice. The 2,4,6-trimethylpyridine (TMP) standard coelutes with 2-pentylfuran, a
lipid oxidation product. The co-elution does not hinder the quantitation of TMP,
because the target ion, at m/z 121, is not found in the fragmentation pattern of
2-pentylfuran.
4. Waxy Rice
Glutinous or waxy rice contains little or no amylose. The rice has a distinctive
aroma and is favored by the Japanese for making traditional rice cakes and crack-
244 Grimm et al.
ers. Mochiminori is a waxy rice and has less than 1% amylose, Koshihikari is a
low-amylose rice (15%), and Nipponbare is a medium-amylose rice (20%).
The Koshihikari variety is the favorite eating rice of the Japanese. It has low
amounts of the unsaturated aldehydes and typically exhibits a lower amount of
volatile compounds in general. Figure 4 shows the relative amounts of selected
unsaturated aldehydes for each of these varieties of rice. Aldehydes typically
possess a slightly green or grassy odor; unsaturated aldehydes possess a fatty or
waxy aroma. All of these compounds may contribute to the grainy/starchy aroma
characteristic of waxy rice. No one unique compound was identified in the waxy
rice GC profile that was not also observed in the nonwaxy rice profile.
IV. QUANTITATIVE ANALYSIS
The complex equilibria established during the SPME process make absolute
quantitation extremely difficult. Most of the literature employing SPME has
been on qualitative analysis. However, quantitation in food products is becoming
FIGURE 4 Comparison of the relative intensities of selected compounds between a waxy

(Mochiminori), a low-amylose (Koshihikari), and a high-amylose (Nipponbare) rice.
more and more prevalent (2426). For liquid samples, the use of standards and
the development of a calibration curve works well. However, for solid samples,
the method of standard additions is generally employed.
A. Recovery
Depending on the compound, the fiber employed, adsorption conditions, and the
matrix, specific analyte recoveries can be as high as 90%. However, when a solid
sample is sampled, agitation and salting techniques are not possible and analyte
recoveries are typically much lower. Employing SPME, for example, the recov-
ery of 2-AP from rice has been determined to be 0.3% (22). Because of this low
recovery, there is a large error associated with absolute concentrations of 2-AP
in rice determined by SPME analysis. However, the data are sufficient to permit
the screening of experimental rice varieties for the presence of 2-AP, and varieties
can be ranked on the amounts of 2-AP present.
FIGURE 5 Observed amounts of 2-AP recovered from 0.75 g of rice with the addition
of various amounts of water. A standard solution containing 2 ng of TMP was added to
each sample.
246 Grimm et al.
B. Internal Standards and Calibration Curves

When calibration curves for SPME analysis are prepared, the standard should be
in a matrix identical to the sample. Slight deviations in the volatile composition
can shift the equilibrium, resulting in irreproducible data. The addition of stan-
dards in organic solutions should be avoided if at all possible. Even the addition
of as little as 1 l of an organic solvent can result in a large decrease in the
amount of the standard adsorbed by the SPME fiber. Figure 5 shows the effect
of the addition of water on the recovery of 2-AP and TMP. Using a 15-minute
adsorption period at a temperature of 65C, a Carboxen/DVB/PDMS fiber was
used to collect the volatile compounds from the headspace of 0.75 g of white
rice. TMP was employed as the internal standard. A 2-l aliquot of a 1 ppm
aqueous solution was added to each vial, thus effectively placing 2 ng of TMP
in each sample. With the successive addition of water to the sample, there was
a decrease in the amount of TMP recovered.
V. SUMMARY
This chapter describes the application of SPME to develop a rapid, sensitive

analytical technique for analyzing the volatile compounds found in the headspace
of rice. A key requirement of this methodology is the heating of the sample in
order to produce sufficient amounts of analytes to be successfully analyzed. Anal-
ysis of less than 1-g samples of milled or brown rice kernels is possible. The
key odorant in fragrant rice is 2-AP, and recoveries are enhanced by the addition
of water. However, the recovery of other compounds may be suppressed by the
addition of water, as observed with the internal standard TMP. This method is
readily amenable for the analysis of additional compounds once they have been
identified as having an impact on the sensory quality of rice.
ACKNOWLEDGMENTS
The authors would like to thank Robert Shiery at Supelco for the discussion on
contaminants originating from the use of SPME.
REFERENCES
1. J. G. Wilkes, E. D. Conte, K. Yongkyoung, M. Holcomb, J. B. Sutherland, and

D. W. Miller. Sample preparation for the analysis of flavors and off-flavors in foods,
J. Chrom. A 880:333 (2000).
2. M. Zhou, K. Robards, M. Glennie-Holmes, and S. Helliwell, Analysis of volatile

compounds and their contribution to flavor in cereals, J. Ag. Food Chem. 47(16):
39413953 (1999).
3. R. J. Stevenson, X. D. Chen, and O. E. Mills, Modern analyses and binding studies
of flavour volatiles with particular reference to dairy protein products, Food Research
International 29(34):265290 (1996).
4. R. Belardi and J. Pawliszyn, The application of chemically modified fused silica
to capillary columns, Wat. Pollut. Res. J. Canada. 24(1):179191 (1989).
5. R. Eisert and K. Levsen, Solid-phase microextraction coupled to gas chromatogra-
phy: a new method for the analysis of organics in water, J. Chromatogr. A 733
(1/2):143157 (1996).
6. H. Kataoka, H. L. Lord, and J. Pawliszyn, Applications of solid-phase microextrac-
tion in food analysis. J. Chromatogr. A. 880:3562 (2000).
7. T. Tsugita, Aroma of cooked rice, Food Rev. Int. 1(3), 497520 (1986).
8. R. Buttery, J. Turnbaugh, and L. Ling, Contributions of volatiles to rice aroma, J.
Agr. Food Chem. 36:10061009 (1988).
9. R. I. Ory and G. J. Flick, Off-flavors in rice and rice products, Off-flavors in Foods
and Beverages (G. Charamblambous, ed.), Elsiever, New York, 1992.
10. J. Piggott, W. Morison, and J. Clyne, Changes in lipids and in sensory attributes on
storage of rice milled to different stages, Int. J. Food Sci. & Tech. 26:615628
(1991).
11. Y. Suzuki, K. Ise, L. Cheng, I. Honda, Y. Iwai, U. Matsukura, and C. Li, Volatile
components in stored rice [Oryza sativa (L.)] of varieties with and without lipoxy-
genase-3 in seeds, J. Agric. Food Chem. 47(3):11191124 (1999).
12. N. J. Yau and T. T. Liu, Instrumental and sensory analysis of volatile aroma of
cooked rice, J. Sensory Sci. 14(2):209233 (1999).
13. R. Widjaja, J. Craske, and M. Wootton, Changes in volatile components of paddy,
brown, and white fragrant rice during storage. J. Sci. Food Agric. 71:218224
(1996).
14. I. Yajima, T. Yani, M. Nakmura, H. Sakakibara, and K. Hayashi, Volatile flavor
components of cooked Kaorimai (scented rice, O. sativa japonica), Agric. Biol.
Chem. 43(12):24252429 (1979).
15. R. Buttery, L. Ling, and B. Juliano, 2-Acetyl-1-pyrroline: an important aroma com-
pound in cooked rice, Chem. Ind. (London), 4 Dec., 958 (1982).
16. M. Petrov, M. Danzart, P. Giampaoli, J. Faure, and H. Richard, Rice aroma analysis:
discrimination between a scented and a non-scented rice, Sciences Des Aliments.
16:347360: (1996).
17. R. Buttery, J. Turnbaugh, and L. Ling, Contributions of volatiles to rice aroma. J.
Agr. Food Chem. 36:10061009 (1988).
18. C. Grimm, E. Champagne, K. Bett-Garber, and K. Ohtsubo, The volatile composi-
tion of waxy rice. Proceedings of the United States-Japan Cooperative Program in
Natural Resources (UJNR) 29th Protein Resources Panel Meeting, Honolulu,
Hawaii, November 1925, 2000.
19. J. Pawliszyn, Solid Phase Microextraction: Theory and Practice, Wiley-VCH, New
York, 1997.
248 Grimm et al.
20. C. Grimm, E. Champagne, K. Bett, Composition of the headspace above selected

rice varieties. Proceedings of the United States-Japan Cooperative Program in Natu-
ral Resources (UJNR) 27th Protein Resources Panel Meeting, Honolulu, Hawaii,
October 2531, J1-J5, 1998.
21. E. T. Champagne and C. C. Grimm, Stabilization of brown rice products using etha-
nol vapors as an antioxidant delivery system, Cereal Chem. 72:255258 (1995).
22. C. Grimm, C. Bergman, J. Delgado, and R. Bryant, Screening for 2-acetyl-1-pyrro-
line in the headspace of rice using SPME/GC/MS, J. Agr. Food Chem. 49(1):245
249 (2001).
23. C. Bergman, J. Delgado, R. Bryant, C. Grimm, K. Cadwallder, and B. Webb, A rapid
gas chromatographic technique for quantifying 2-acetyl-1-pyrroline and hexanal in
rice (Oryza sativa, L.) Sub. to Cereal Chemistry 2000.
24. B. D. Page and G. Lacroix, Application of solid-phase microextraction to the head-
space gas chromatographic analysis of halogenated volatiles in selected foods. J.
Chrom. 648:199211 (1993).
25. S. Lloyd and C. Grimm, Analysis of 2-methylisoborneol and geosmin in catfish by
microwave distillation/solid-phase micro-extraction, J. Agric. Food Chem. 47(1):
164169 (1999).
26. Z. Zhang and J. Pawliszyn, Quantitative extraction using an internally cooled solid
phase microextraction device, Anal. Chem. 67(1):3443 (1995).
9
Headspace Techniques for the
Reconstitution of Flower Scents
and Identification of New
Aroma Chemicals
Thomas McGee and Kenneth L. Purzycki

Givaudan Fragrances Corp., Teaneck, New Jersey
I. INTRODUCTION
Until the mid-nineteenth century, the only raw materials available to perfumers
for creating fragrances were the extracts of plants and a few animal materials
such as musk and civet. The most important raw materials were the flower scents.
It is not surprising that Grasse, the center of the fragrance industry in the nine-
teenth century, was also the center of flower cultivation in France. The scent of
the flower is from the so-called essential oils. These oils are composed of mole-
cules that have a molecular weight below 400 and sufficient vapor pressure at
ambient temperature to volatilize. The essential oils can occur in several parts
of the plant and their usefulness to the plant is varied. For example, flowers use
scent to attract insects to pollinate them, and oils found in the roots and barks
of plants often have antimicrobial properties, thought to be part of the plants
defense mechanism.
The essential oils of flowers for perfume creation were obtained by a variety
of methods, such as steam distillation, solvent extraction, and enfleurage (1).
Steam distillation is the least expensive process. The flowers are mixed with
water and steam and the essential oils are then distilled off. The distillate separates
into an oil and water phase. The heat required, however, can chemically modify
some of the components, and the more water soluble ones can stay in the water
phase. Solvent extraction is used with fresh flowers; water-immiscible solvents
249
250 McGee and Purzycki
are used because flowers have a water content of 8090%. The solvent is removed
under vacuum. For some flowers, such as jasmine and tuberose, yield of essential
oils by solvent extraction is low. The yield can be increased by exploiting the
property of some flowers to continue to emanate scent after picking. The tech-
nique is called enfleurage, in which freshly picked flowers are layered onto puri-
fied fat into which the scent continues to migrate. The essential oils are recovered
from the fat by solvent extraction or steam distillation. The production of the
essential oils of flowers is expensive and the supply and quality depend on the
harvest. This severely limited the size of the early fragrance industry.
The beginning of the nineteenth century saw chemists starting to synthesize
the key aroma chemicals. The leading chemists of the era, including five Nobel
prize winners, were associated with the synthesis of aroma chemicals (2). The
chemicals synthesized in this period were the breakthroughs responsible for creat-
ing a new aroma chemical industry. This provided synthetic aroma chemicals at
dramatically lower costs and their quantity and quality was not affected by har-
vests. Perfumers were able to create scents that could be supplied in sufficient
amounts and at a cost affordable by the general population, which created a grow-
ing demand. The fragrance industry is the size it is today only because of the
success of chemists in unlocking the secrets of natural aromas. Research on
flower scents continued to lead to the discovery of new aroma chemicals (3).
The introduction of gas liquid chromatography (GC) in the early 1950s
accelerated the discovery of new aroma chemicals. The GC gave fragrance scien-
tists a powerful technique to separate components from the very complex mixture
that makes up a flower scent. In addition, it allowed chemists to collect aroma
chemicals separated by GC in quantities that enabled them to be easily identified
by mass spectrometry (MS) (4). It was only natural that these instruments were
later coupled together. The combination of these two techniques, referred to as
GC/MS, now provides the most sensitive analytical technique for separating and
identifying flower scent volatiles.
The majority of the early work focused on investigating natural extracts.
However, techniques to obtain the true character of flower scents still held high
interest for the chemist and the perfumer. The extracts of flowers often lack the
delicate aroma of a flower at the peak of its life cycle. Some extractive techniques
require heat, which often eliminates key volatile fragrance components and can
generate artifacts by hydrolysis, oxidation, or breakdown of the plant metabolites.
Flower scents of especially rare species (which cannot be picked) or flowers that
are not found in sufficient quantity for extraction are also of interest as a source of
new aroma chemicals and scents. This chapter concentrates on the nondestructive
headspace techniques that have been developed for the collection of flower scents.
The techniques and discussion also have relevance for the collection and identifi-
cation of aroma chemicals in the headspace of food and beverage samples, as
well as many other types of samples.
HS Techniques, Reconstitution, and Identification 251
II. THE SCENT OF FLOWERS
An appreciation of the genesis and function of a flowers scent is important to

the fragrance scientist in deciding which flower to study and also in planning a
strategy to capture it. The majority of volatile components responsible for the
flowers scent are metabolites produced by enzymatic processes (5), which are
stored in specialized oil glands. For example, Figs. 1 and 2 show biochemical
pathways that produce the most abundant volatile chemicals found in nature.
Figure 1 illustrates the enzymatic pathway that generates isoprene, a basic build-
ing block for a large variety of acyclic and monocyclic terpenes such as myrcene,
ocimene, and limonene. Figure 2 shows how isoprene can also be enzymatically
converted to acyclic, oxygenated monoterpenes such as linalool, geraniol, and
alpha terpineol. These biochemical processes can be influenced by many factors.
The primary function of a flowers scent is to attract a pollinator. Flowers
and insects have coevolved, and the biochemical processes determining the type
and quantity of aroma chemical in the flowers scent have evolved to maximize
the flowers attraction to its pollinators. This can result in new molecules or new
scents that are specific to a pollinator or class of pollinator. Observation of the
pollinators and their activity can provide the chemist with the first insight into
which flowers to study.
Plants have to expend energy to produce scent and, therefore, maximum
output and optimum quality of scent is produced only during the period when
the insects that pollinate them are present. The pollinators abundance gives the
scientist information on the best time to collect the scent of a flower. Moth-
pollinated flowers, for example, are completely scentless during the day but ema-
FIGURE 1 Enzymatic pathway to isoprene.

FIGURE 2 Overview of the pathway leading to monoterpene formation in plants.
nate a powerful scent during the night to attract night moths (6). Conversely,
bee-pollinated flowers issue their scent only during the day. The exact nature of
the controlling mechanism for scent emanation is not fully understood. It is
thought that factors such as light intensity and temperature regulate the plants
internal biological clock.
Other factors such as soil, climate, position relative to the sun, and so forth
also have a major role in the scent of the flower. Hence, collection strategies
must be designed to take into account factors such as biorhythm, maturity, and
the need to collect flower scents in their indigenous location.
III. HEADSPACE TECHNIQUES
Headspace techniques can be classified into two types: static and dynamic (7).
Static headspace encloses the flower for a set period of time, and volatiles col-
lected at the end of the period are analyzed. Dynamic headspace essentially
moves the aroma chemicals continuously from the flower directly to the collect-
ing matrix.
The first headspace technique used to collect the scent of a flower was a
static technique developed by Dodson and Hill in 1966 (8), who were trying to
identify the aroma chemicals responsible for attracting Euglossine bees to several
species of orchids. They placed the orchid in a sealed jar. After 30 minutes, a
gaseous sample was removed with a syringe and injected directly into a GC.
They were able to identify the different components of the orchids scent using
this headspace analysis technique. Their simple experiments started a new path of
research that ultimately transformed the approach of fragrance scientists studying
flower scents.
Static headspace, however, has drawbacks for the collection of flower
scents. It requires that the sample be collected over a relatively long period,
whichbecause of the biorhythm of flowersprovides an average scent compo-
sition that is not the optimal scent. Static collection also yields very low concen-
trations of the less volatile aroma chemicals, often below the detection limits of
the analytical method.
Dynamic headspace is capable of collecting enough scent in a relatively
short time for quantitative analysis. The flower to be studied is placed into a
headspace collection device to minimize airborne environmental contaminates.
The collection containers are commonly constructed of glass and contain provis-
ions for an entrance port, an exit port, and an opening that will allow the device
to be placed around or over the flower without damaging it. The size and shape
of the containing vessel is selected to suit the flower being studied. A typical
dynamic headspace collection device is shown in Fig. 3. The filter at the entrance
port contains a material such as activated charcoal to adsorb airborne contami-
nants. The exit port has a trap containing the material on which the flowers scent
will be adsorbed. This is connected to a suitable pump or vacuum device. The
air sweeps the volatiles emitted from the flower into the collection trap, where
they are trapped on adsorbent material.
One alternative to dynamic headspace is the closed loop headspace col-
lection device, developed by Brunke et al. (9), shown in Fig. 4. This device
places a flower within a collection vessel, and the volatiles are collected on an
appropriate trapping material. The difference between this method and con-
ventional dynamic headspace is that the closed loop technique constantly circu-
lates the air within the device using an in-line pump. Results obtained with the
closed loop system are similar to those obtained with conventional dynamic head-
space.
Dynamic headspace is the preferred technique for collecting flower scents.
Dynamic headspace preconcentrates the volatiles onto an adsorbent trap from
FIGURE 3 Typical device used to collect headspace samples from flowers.
FIGURE 4 Diagram showing a closed-loop headspace collection device. 1, Conical flask;

2, flower; 3, battery-powered suction pressure pump; 4, headspace collection trap; 5, acti-
vated charcoal tube (purification of the air stream); 6, direction of airflow.
which they can be removed by either solvent extraction or thermal desorption.

The advantages of dynamic headspace collection are as follows:
1. It increases the concentration of the aroma chemicals adsorbed onto
the trap. This means that a shorter collection time is required so that
the sample can be taken at the peak of the flowers scent emission.
2. Sufficient sample size is also obtained to enable the less volatile aroma
chemicals to be analyzed.
A. Trapping Methods
1. Adsorption on Porous Material
The choice of the trapping material depends on a number of factors, such as the
presence of moisture, the volatility and functionality of the aroma chemicals, and
the method of desorption. Because all living flowers transpire, giving off water
vapor, the trapping material used must not be deactivated by water and it must
have a high affinity for volatile organic materials with a large range of polarities.
The initial adsorbent of preference was activated charcoal. For example, Lampar-
sky (10), among others, reported studies of flower scents using headspace and
activated charcoal as the adsorbent. Other materials, such as the porous polymer
Porpak Q, based on ethylvinylbenzenedivinylbenzene copolymer, and Tenax ,
based on 2,6-diphenyl-p-phenylene oxide, were developed to replace activated
charcoal because of the difficulty in effecting complete desorption from charcoal,
particularly of polar compounds (11).
Notwithstanding this drawback, activated charcoal continued to be widely
used to study flower scents (12) until 1993, when Brunke et al. (13) highlighted
problems with activated charcoal in the comparative headspace studies of several
different charcoals and Tenax. The design of the study avoided differences due
to the maturity and the biorhythm of the flower by collecting samples from the
same source at the same time, using a flask with multiple ports shown in Fig.
5. The traps containing different adsorbents were pumped at similar rates. The
investigators found that the analysis of the headspace samples adsorbed on char-
coal showed the presence of several octatrienols that were not present in samples
adsorbed on Tenax. They later showed, by collecting the headspace sample of a
mixture of cis and trans ocimene (Fig. 6), that the octatrienols were actually
artifacts caused by the oxidization of ocimene in the presence of charcoal (14).
The use of charcoal for the collection of flower scents is now very limited.
The choice of the porous polymer for flower headspace studies is dependent
on the method of desorption and the primary purpose of the collection. Porpak Q
traps are often employed when solvents are used to remove the collected volatiles.
Normally, a small trap is used so that the sample can be extracted with a minimal
amount of solvent. A custom-made glass trap is shown in Fig. 7. It contains 5 to
FIGURE 5 Closed loop headspace apparatus for the simultaneous headspace analysis of
flower fragrances using different adsorbents.
FIGURE 6 Formation of artefacts related to the collection of headspace of cis/trans-beta-

ocimene using commercial charcoal adsorption tubes. Gas chromatograms (60 m 0.25
mm I.D. DB Wax column) of (A): cis/trans-beta-ocimene (19,21) and (B): Headspace
sample of cis/trans-beta-ocimene collected on commercial charcoal adsorption tubes.
FIGURE 7 Typical solvent desorption headspace trap. (a) Glass capillaries; (b) glass wool;
(c) Porpak Q; (d) microliter syringe; (e) ampoule.
10 mg of Porpak Q (c) held in place by small glass beads (b), which are fixed by
another glass tube (a) fused to the outer glass tube. This small unit is particularly
useful for removing trap material with a minimum amount of solvent (50 to 100
l). Extraction requires attaching a syringe (d) with heat shrink tubing to the top of
the trap and a small metal capillary to the bottom of the trap. The solvent is pushed
slowly through the Porpak Q and collected in a small micro flask (c) as shown. A
potential problem with this system is that it can easily introduce contaminates. The
analyst must take care that the system is clean and the solvents are of the best quality.
An advantage of solvent desorption is that it makes more than one analysis possible,
making it particularly suitable when searching for new aroma chemicals.
Tenax, because of its thermal stability, is normally selected as the adsorbent
when thermal desorption techniques are used. Thermal desorption heats the head-
space trap and transfers the trapped volatiles with an inert gas to a cryofocusing
area, which is later heated to transfer the volatiles directly onto a GC column.
Thermal desorption techniques have gained popularity with the development of
automatic thermal desorption instruments such as the Perkin Elmer ATD-400,
Gerstel TDS/CIS, and Scientific Instrument Services Autodesorb. Automation
makes thermal desorption a suitable method for studying a large number of flower
samples. It also minimizes contamination and avoids sampling errors. The major
drawback to thermal desorption is that only one analysis per sample is possible.
B. Solid-Phase Microextraction (SPME)

SPME is a relatively new trapping technique. First reported by Berlardi and Paw-
liszyn in 1989 to analyze organic materials in a water matrix (15), it was then
developed by Pawliszyn and coworkers as a headspace sampling technique for
capturing volatile components (16). The method concentrates analytes using a
silica fiber coated externally with an adsorbent, such as polydimethylsiloxane.
Analytes adsorbed on the fiber are easily removed by placing the fiber directly
into the heated injection inlet of a GC. The trapped volatiles are thermally de-
sorbed directly into the GC column and analyzed. It is an increasingly popular
technique for sampling volatiles in a variety of matrices.
Mookherjee et al. (17) reported the first use of SPME as a trapping tech-
nique for flower scents. They claimed, in their study of the scent of the orchid
flower, Dendrumbium superbum, that it was a superior technique to classical
headspace. However, they reported no direct comparison to dynamic headspace
techniques, and other studies indicate that SPME might not be a suitable tech-
nique for the quantitative determination of the components in a flowers scent.
Yang and Peppard (18), for example, showed marked differences between SPME
and Tenax in their study of volatiles in ground coffee. The conventional head-
space technique was more sensitive to molecules with a high vapor pressure than
SPME, but SPME extracted more of the less volatile analytes. Similar conclu-
sions reached by Pelusio et al. (19) and Miller and Stuart (20) were that SPME
was less suited for quantitative analyses than gas-sampled static or dynamic head-
space. Elmore et al. (21) not only supported these findings but also reported that
SPME gave a greater number of artifact peaks than a Tenax trap. In contrast,
work by Krubein and Ulrich (22) indicated that SPME could be considered for
collecting flower scents as they concluded that SPME and Tenax were in principle
equal for characterizing aroma volatiles.
McGee and Purzycki (23) reported the first comparative study of SPME
and dynamic gas-sampled headspace for collecting flower scents. They compared
different SPME polymers to dynamic headspace using Tenax and agreed with
the conclusions in the majority of the literature. Each SPME polymer had a
marked bias for the polarity of the aroma chemicals similar to the polymers
polarity, indicating that SPME is not ideal for collecting flower scents. Purzycki
and Sgaramella (24) refined this study by optimizing the collection conditions
for SPME. The optimized conditions for SPME were a contact time of 30 minutes
and a close location (1 cm) to the flowers scent. The relatively long contact time
and close approximation to the flower is necessary to ensure sufficient parti-
tioning of the analytes takes place into the solid phase for analysis. The scent of
a hyacinth flower was collected on the different commercial SPME fibers using
the optimized conditions. A 300-ml sample was also collected by dynamic head-
space on a trap containing 100 mg of Tenax using a pumping rate that allowed the
collection to be done in 30 minutes. All samples were collected simultaneously to
avoid differences due to the factors discussed in Section 4. The volatiles collected
on the Tenax trap and SPME fibers were thermally desorbed and analyzed. The
results are shown in Table 1.
TABLE 1 Headspace Compositions of a Blue Hyacinth Collected with Different

Commercially Available Solid-Phase Microextraction (SPME) Fibers
SPME Fibers
Components Tenax Gray White Blue Black Red Orange
Benzaldehyde 1.80 0.55

Myrcene 2.51 2.96 1.12 7.09 2.57
alpha-Pinene 1.35 1.11
cis-Ocimene 0.40 1.15 0.93
Limonene 0.72 0.75 0.22 0.81
Benzyl alcohol 2.29 1.34 1.67 0.92 3.01 0.58
para cymene 2.63
trans-Ocimene 21.58 21.78 2.37 15.01 10.38 5.16 20.15
Phenylacetaldehyde 1.51 3.35 0.71 1.88 4.33 0.37 3.07
Acetophenone 0.34
Methyl benzoate 0.20 0.24 0.20 0.37
Phenylethanol 28.91 30.85 25.2 23.78 36.06 20.82 26.87
Allo-ocimene 1.99 4.86 0.72
Benzyl acetate 16.89 18.92 13.88 20.15 23.34 15.2 17.45
Methylphenyl acetate 0.54 0.69 0.25 0.78 0.70 0.81 0.87
Aldehyde C-10 0.22 1.28 1.22
Phenylpropyl alcohol 0.23 0.18 0.41 0.2 0.25 0.23
Phenylethyl acetate 2.96 2.66 3.01 3.34 2.28 4.51 2.76
Cinnamic aldehyde 0.68 0.44 0.60 0.32 0.58
Cinnamic alcohol 1.94 1.25 5.08 2.96 0.33 4.04 1.82
Methyl-2-methoxybenzoate 0.24 0.18 0.6 2.64
1,2,4-Trimethoxy benzene 5.71 4.65 12.05 8.30 3.37 13.81 5.88
4-Methoxy phenylethyl alcohol 0.36 0.40 0.39 0.80 0.43
Methyl eugenol 0.38 0.49 1.14 0.81 1.40
Cinnamyl acetate 0.26 0.23 0.75 0.40 0.69 0.30
alpha-Farnesene 3.90 2.68 7.85 8.15 13.89 3.35
Benzophenone 9.88
Benzyl benzoate 4.29 2.38 12.53 8.53 0.46 13.22 5.76
Phenylethyl benzoate 0.65 0.40 2.42 1.41 1.99 1.03
The analytical results from the different SPME fibers tested were, indeed,
highly variable. The selectivity of the adsorption of flower scent analytes with
a different polarity can be significantly altered through the use of different solid
phases on the fiber. The SPME fiber DVB/Caboxen/PDMS obtained the chemical
profile of the hyacinth scent most similar to that collected by Tenax. The research-
ers concluded that SPME does not offer an advantage over dynamic headspace
methods of collecting headspace samples on porous polymers such as Tenax.
The efficiency of collection of the SPME technique depends on both the polarity
of the fiber and the polarity of the aroma chemicals. The main disadvantages of
SPME are that the fibers are fragile and that due to their higher specificity, more
than one fiber may be required per flower analysis to collect all the aroma chemi-
cals in a flowers scent. SPME can be considered a complementary technique to
dynamic headspace gas collection for the study of specific analytes in a flowers
scent.
C. Zenith Trap
McGee and Purzycki (25) developed a new collection method to overcome the
problems of using SPME. The basis of the system is fused silica GC capillary
tubes. The capillary tubes have their inner surfaces coated by adsorbent polymers.
To ensure the adsorption of a wide range of aroma chemicals, several tubes coated
with polymers of different polarity, respectively, were made into a bundle. This
was called the Zenith trap. The size of the bundle was made to fit into a thermal
desorber. The capillary bundle is held at the end of a PTFE heat-shrinkable con-
necting tube in an airtight manner (Fig. 8). The connecting tube links the Zenith
trap to a small pump so that the flower scent volatiles can be drawn through the
capillary bundle and adsorbed on to the tube coatings.
The Zenith trap constructed from GC capillary tubes obtained from Supelco
(26) was directly compared to conventional dynamic headspace with a Tenax
trap by simultaneous collection of the scent of a blue hyacinth flower. The results
obtained from the bundled capillaries are compared to the Tenax headspace trap-
ping system in Table 2. It was concluded that the Zenith trap is comparable to
FIGURE 8 Zenith trap.

TABLE 2 Comparison of Headspace Composition

of a Blue Hyacinth Collected on Both Zenith and
Tenax traps
Components (%)
Component Zenith Headspace
Benzaldehyde 1.2 2.3
p-Cymene 0.2 0.2
Limonene 0.2 1.2
Benzyl alcohol 1.5 2.5
Phenylacetaldehyde 1.0 1.1
trans-Ocimene 1.0 7.2
Phenylethanol 25.5 22.2
Benzyl acetate 13.7 20.1
Aldehyde C-10 0.6 0.5
Phenylpropyl alcohol 1.4 0.7
Phenylethyl acetate 2.4 7.7
Cinnamic aldehyde 4.6 2.9
Cinnamic alcohol 16.0 9.0
Methyl 2-methoxybenzoate 0.7 0.8
Eugenol 0.6 0.8
1,2,4-Trimethoxy benzene 11.6 7.2
Methyl eugenol 0.6 0.9
Cinnamyl acetate 0.5 0.2
alpha-Farnesene 6.3 5.1
Benzyl benzoate 9.3 6.3
Phenylethyl benzoate 1.1 1.1
100.0 100.0
Tenax and would be able to collect aroma chemicals with a wide range of polari-
ties. Figure 9 shows the effect of collection time on the aroma chemicals col-
lected. The Zenith trap is capable of collecting sufficient quantities of headspace
material in around 5 minutes for quantitative GC/MS analysis. The Zenith trap
combines the advantages of SPME and dynamic headspace. It overcomes the
problems of fiber fragility, long extraction times, and the need to analyze and
combine the results of several SPME fibers.
IV. PRACTICAL ASPECTS OF FLOWER SCENT ANALYSIS

A. Living Versus Cut Flowers
Mookherjee et al. (27) reported differences in the composition of headspace sam-
ples collected on Tenax between living flowers and cut flowers. They showed
FIGURE 9 Gas chromatograms (60 m 0.25 mm I.D. SPB-1 MS column) representing

the headspace collected with a Zenith trap at different times. (A) 2 min. adsorption; (B)
5 min. adsorption; (C) 10 min. adsorption.
that the headspace composition of a picked flower had a higher concentration of

cis-3-hexenyl acetate compared to the levels obtained from the same flower still
growing. They reported, therefore, that the scent reconstitution based on the head-
space of picked flowers would not be representative of the true aroma of flowers
still growing.
Kaiser (28) repeated the experiment using a hybrid tea rose. Headspace
samples were collected from two flowers of identical maturity. One rose was left
attached to the bush; the second was cut and placed into water. The headspace
compositions of the two flowers were essentially identical. Kaiser continued to
take headspace samples throughout the day and found no differences in the con-
centration of cis-3-hexenyl acetate for the flower attached to the bush and the
picked flowers. Kaisers study concluded that if cut flowers are used for head-
space studies, the sample must be collected from properly nourished cut flowers.
If this is done, then the results are similar to those obtained from flowers still
attached to the plant.
B. Maturity of the Flower

Flowers go through a maturing process and the speed of maturation depends on
the flower. The effect of flower maturity on the headspace composition of the
flowers scent is illustrated in Fig. 10. This shows some of the major components
of the headspace of a hybrid Asian lily called Gigi on the first day of the bud
FIGURE 10 Headspace analysis of Asian lily Gigi at different times in the blooming
cycle. The area chart represents the total nanograms per liter at 0, 1, 24, and 72 hours are
shown on the right scale. The bar chart represents the nanograms per liter of the component
shown.
opening and on day 2 and day 4, respectively. Collections were made at the same
time of day to obviate biorhythm effects. The concentration of aroma chemicals
(nanograms per liter) is highest on the first day of the flower opening. The total
nanograms per liter decreases from approximately 10,000 on day one to around
2,000 on the fourth day after blooming. The scent of the flower can be seen to
change as the flower matures. For example the ratio of linalool to limonene
changes from 4 :1 on day 1 to 6 : 1 on day 4. This study clearly shows that the
quality and quantity of the lilys scent is markedly affected by the flowers matu-
rity and is optimal when the flower is first fully open.
C. Biorhythm
The flowers biorhythm geared to attract pollinators has been discussed. Its im-
portance to scent collection can be seen in Fig. 11. This shows the total amount
of volatiles and four of the major components of the headspace of Brugmansia
sauviens collected over a 24-hour period. The biorhythm of the flower can be
seen from the total amount of volatiles that the flower emits. The flowers scent
production starts to increase markedly after 7:00 p.m. The concentration of vola-
tiles in the headspace increases from around 2000 nanograms per liter to over
6000 nanograms per liter after 11:00 p.m., peaking at over 7000 nanograms per
liter between 2:00 a.m. and 3:00 a.m. The quality of the scent can also be seen
to change. Nerolidol, for example, is present in the headspace only between
9:00 p.m. and 5:00 a.m. The ratios of the components and, hence, the scent of
the flower change over the 24-hour period. The quantity and quality of the flow-
FIGURE 11 Biorhythm chart of Brugmansia sauviens. The area charts depict the nano-
grams per liter of the component at different times over a 24-hour period.
ers scent is tailored to match the maximum pollinator activity. It is very impor-
tant for the scientist to determine the time of the flowers peak scent emission
for its collection and analysis.
D. Location of Flower
Soil, topography, and climate not only influence the quality of wine but can also
greatly affect the quality of a flowers scent. The biochemical processes that pro-
duce essential oils have been shown to be markedly affected by conditions where
the flower is grown (29). This study showed that factors such as light, temperature,
moisture, and nutrition affect the metabolic processes that produce the essential
oils. This must be taken into consideration in scent collection strategies and com-
parative studies. The strong influence of environment on the flowers scent is the
main reason why scientists should study flowers in their natural habitat (30).
In recent years, the search for new scents and aroma chemicals has led
fragrance scientists to extend their exploration of flowers and plants to the rain
forests. Rain forests exist in many tropical regions and they are home to the major-
ity of flowering plants. Many of these rain forest flowers have never been studied
and they are in danger of disappearing as the rain forests are destroyed. The prom-
ise of this undiscovered cornucopia of flower scents and new aroma chemicals
gave the initial impetus to carry out scent exploration in the rain forests (31).
1. Automated Headspace Sampler

When studies are carried out in botanical gardens or greenhouses, it is possible
to observe the pollinators or to smell the flower scent at various times to assess
the flowers biorhythm. However, in locations such as the rain forest this may
not be practical, especially for night-scented flowers. To collect and identify sam-
ples at the peak of a flowers biorhythm, an automated collection system was
developed (30). This is shown in Fig. 12. The portable sampling device consists
of a manifold containing 12 headspace traps. Each trap is connected via a micro-
valve to a pump. A computer is used to switch the valves open, turn on the pump,
and control the duration of the collection. The scientist typically knows if the
flower emits its scent during the day or night, and each trap will be set to collect
a 1 hour sample over 12 hours. Subsequent analysis identifies the collection pe-
riod at which the plant was producing its maximum scent. The analytical data
of this headspace sample is used to develop a skeletal formula for the reconstitu-
tion of the flowers scent.
2. Flower Scent Collection in Rain Forest Canopies

The canopy of the rain forest is the last unexplored world of plants. Botanists
have been developing methods for approaching this final frontier (32) using tree-
FIGURE 12 Schematic drawing of an automated headspace collection apparatus used to

collect flower biorhythm data.
climbing spikes, cantilever devices, hot air balloons, and so forth. Using one of
these techniques to gain access to the canopy, the fragrance scientist still has the
challenge of capturing the flowers scent. The fragile nature of the canopy may
require locating and securing equipment on a slender network of branches. In
addition, flowers can be located at the extremity of branches, on which headspace
devices cannot be positioned or suspended.
Lightweight equipment has been specially designed to sample flower scents
in the canopy. The lightweight system is a miniaturization and redesign of the
automatic collection device described in Sec. IV.D.1. This smaller version has
the weight reduced by a factor of 20. Although glass headspace collection vessels
are easily cleaned and reusable, they are fragile, bulky, and can be too heavy
to suspend on fragile branches. Plastic bags are a simple alternative. They are
commercially available in a variety of sizes. They are lightweight, nonbreakable,
and disposable, eliminating the need to clean.
Ideally, a plastic bag should not contain volatile materials that could mask
fragrance ingredients. A large variety of commercially available plastic bags have
been investigated, including Tedlar plastic bags (33), which are specially de-
signed to collect environmental gas samples. The headspace analysis of a repre-
sentative sandwich type plastic bag, roaster type plastic bag, and a Tedlar bag
are shown in Figs. 13a, b, and c, respectively. In general, the volatiles emitted
FIGURE 13 Gas chromatograms (60 m 0.25 mm I.D. SPB-1 MS column) depicting

the volatiles emitted by different plastic bags. (A) Typical sandwich plastic bag; (B) micro-
wave roaster bag; (C) Tedlar bag.
from plastic bags are composed of unpolymerized monomers, plasticisers, and

chemicals associated with their manufacture. The type and quantity depend on
the plastic used. The bags with the lowest volatiles are the roaster plastic bags
designed for cooking foods and Tedlar bags, respectively. The major component
of the headspace of roaster bag is caprolactam, a component associated with
the manufacture of nylon. The major component in the headspace of the Tedlar
environmental collection bag is phenol.
The concentration of the volatiles emitted from a plastic bag was also found
to increase as the ambient temperature increased. Figure 14 shows the total con-
centration of volatiles emitted from different plastic bags at 25C and 45C, re-
spectively. From this study, roaster bags are recommended for use as headspace
collection vessels as they emit relatively few volatiles even at the higher tempera-
ture. A more important criteria for selection of the roaster bag is that the major
FIGURE 14 Total concentration of volatiles emitted from plastic bags at (A) 25C and
(B) 45C.
contaminant, caprolactam, cannot be confused with a naturally occurring sub-

stance.
The size of the plastic bag can be tailored to the flower. The bag is placed
over the flower and a headspace trap may be simply placed in the open end,
which is then sealed around the stem of the flower with tape or a clamp. Care
has to be taken to prevent creating a vacuum during collection. To minimize this,
the internal volume of the bag should be two to three times larger than the size
of the flower sample to be collected. A better and more convenient method is to
attach a plastic fitting near the top of the plastic bag. The plastic connector is
attached to the bag by making a hole in the bag and attaching it with a nut and
a washer. If the bag is thin, the area around the hole can be reinforced with
adhesive tape. The other end of the connector has a self-sealing nut, which accepts
a headspace trap with an outer diameter of about 0.635 cm and seals the head-
space trap into the bag (Fig. 15). This configuration allows the headspace trap
to be placed closer to the center of a flower, and permits a filter charcoal trap to
FIGURE 15 Typical setup showing the collection of volatiles from a flower using a
plastic bag.
FIGURE 16 Headspace collection using a Zenith trap and support apparatus.
be placed on the stem and sealed as above. This setup eliminates the probability
of creating a vacuum and allows smaller bags to be used.
To reach a flower at the tip of a slender branch and collect its scent, a
system was designed in which the Zenith trap discussed in Section III.C was
incorporated in an extendable lightweight collection rod that can reach out to
flowers on the outer tips of canopy branches. The Zenith trap is housed in a
sheath attached to an extending aluminum rod. On reaching the flower, the Zenith
trap is extended from its protective sheath, and the scent of the flower drawn
though it by a small pump for between 5 and 10 minutes, at a pumping rate of
between 40 and 60 ml of gas per minute as shown in Fig. 16. The flower scent
volatiles are trapped. After collection the Zenith trap is removed and stored. The
volatiles are subsequently removed from the absorbent capillaries in the labora-
tory, using a thermal desorber system, and analyzed by GC/MS.
V. APPLICATION OF HEADSPACE DATA

A. Reconstitution of the Flower Scent
The challenge for the fragrance scientist is to create a liquid formula that will
reproduce a flowers scent in the vapor phase. Headspace analysis provides the
fragrance scientist with a qualitative and quantitative picture of the components

present in the air surrounding the odor-emanating source. Translating this to a
liquid formula based solely on the distribution of chemicals present in the head-
space does not provide an accurate reconstitution. Physical chemistry laws teach
us that high vapor pressure components will be proportionately higher in the
vapor phase than they are in the liquid phase and, conversely, those with low
vapor pressure will be lower. Converting the headspace formula to a liquid for-
mula is also complicated because the headspace is collected in an open system
rather than a closed systemi.e., it is a non-equilibrium system. At present there
is no simple way to convert exactly non-ideal gas compositions into liquid com-
positions. It is possible, however, for the fragrance scientist to derive from the
headspace analysis a skeletal liquid formula that approximately reproduces the
scent of the flower. Using the following equations, the vapor pressure contribu-
tion of each component is estimated (Eq. 1) and the percentage each individual
vapor pressure is of the total vapor pressure of all the components is determined
(Eq. 2).
P z (Wt z ) (MW z )/(Vp z ) (1)
% Zl (P z ) 100/TVP (2)
where:
Wt z Weight of component Z in the headspace weight.
MW z Molecular weight of component Z.
Vp z Vapor pressure of component Z.
Pz Vapor phase contribution of component Z
Zl Component Z in the liquid phase
TVP Total vapor pressure of all components.
Vapor pressures can be either measured (34) or calculated based on the molecular
structures using existing software (35). This approach is demonstrated in Table
3, which shows data collected from the headspace of a night-blooming cactus
flower, Epiphyllum oxypetalum, and the reconstituted liquid formula.
The first reconstitution using computed partial vapor pressures more closely
resembles the scent of the flower than a simple reconstitution of the headspace
composition. An exact replication can be achieved by the fragrance scientist who
has perfumery training, or is working with a perfumer, to fine-tune the liquid
formula.
Other practical aspects must be taken into account when the headspace
reconstitution is for use in a commercial fragrance. It is possible to simplify the
liquid formula by eliminating all components that do not contribute to the flowers
scent (i.e., those components that occur in the headspace below their perception
threshold). Perception thresholds of aroma chemicals can be obtained from data-
bases. If a database is not available to the scientist, the olfactive profile of the
TABLE 3 Skeletal Liquid Formulation Based on the Headspace Analysis
of Epiphyllium oxypetalum
Epiphyllium Oxypetalum Percent
Rt Material MW MM Headspace Liquid

7.65 n-Hexanal 100.0 10.8880 0.02 0.0000
12.44 trans-Allocimene 136.0 0.9690 0.49 0.0011
13.08 Benzaldehyde 106.0 0.9740 1.15 0.0019
14.33 p-Mentha-1(7),2,8-triene 134.0 1.5350 0.36 0.0005
14.49 6-Methyl-5-hepten-2-one 126.0 1.2770 0.13 0.0002
14.72 Phenol 94.0 0.6140 0.05 0.0001
14.83 alpha-Myrcene 136.0 2.2900 0.30 0.0003
15.14 1,3,5-Trimethyl benzene 120.0 2.3240 0.03 0.0000
15.43 Myrcene 136.0 2.2900 5.33 0.0049
15.87 alpha-Phellandrene 136.0 2.2900 0.27 0.0003
16.39 Benzyl alcohol 136.0 1.8560 7.81 0.0088
16.92 Limonene 108.0 0.1580 1.73 0.0183
17.12 cis-B-ocimene 136.0 1.5410 0.08 0.0001
17.37 Melonal 136.0 2.2900 1.93 0.0018
17.76 trans-Ocimene 140.0 0.6220 9.84 0.0341
17.97 alpha-Terpinene 136.0 1.5590 0.53 0.0007
18.83 Methyl benzoate 136.0 1.0750 1.95 0.0038
18.95 Terpinolene 136.0 0.3400 1.35 0.0083
19.37 2-(1Pentenyl)-furan 136.0 1.1260 0.44 0.0008
19.45 Phenylethyl alcohol 136.1 1.3520 0.44 0.0007
20.07 Nonadecane 142.6 1.5830 0.09 0.0001
20.84 allo-Ocimene 136.0 0.9690 3.73 0.0081
21.67 (E)-iso Citral 152.0 0.0710 0.13 0.0044
21.81 Ethyl benzoate 150.0 0.1800 0.13 0.0017
22.98 Methyl salicylate 152.0 0.0700 13.38 0.4478
23.36 n-Decanal 156.0 0.2070 0.88 0.0103
23.97 Dodecane 170.3 0.2090 0.15 0.0018
24.37 Nerol 154.0 0.0130 0.80 0.1465
24.49 Neral 152.0 0.0710 0.79 0.0259
25.34 Geraniol 154.0 0.0130 27.28 4.9807
26.26 2E-6E-Farnesol 222.0 0.0005 0.07 0.4975
26.35 Indole 117.0 0.0300 0.11 0.0065
26.51 2-Oxo-citronellol 170.2 0.0020 1.43 1.8804
26.63 Neryl acetate 258.0 0.0005 0.06 0.4859
27.69 Tetradecane 198.0 0.0040 0.10 0.0753
28.21 2,3-Epoxy geraniol 170.2 0.0070 1.06 0.3972
28.79 Eugenol 164.0 0.0060 0.13 0.0549
29.79 Geranyl acetate 196.0 0.0260 0.51 0.0598
30.03 Benzyl isovalerate 192.0 0.0290 1.20 0.1227
31.82 Isoeugenol 164.0 0.0005 0.52 2.6120
32.13 Geranyl acetone 194.0 0.0020 0.06 0.0906
33.32 Benzyl tiglate 190.2 0.0080 0.22 0.0796
34.84 n-Amyl salicylate 208.0 0.0020 0.05 0.0779
37.16 Geranyl isobutyrate 224.0 0.0020 0.03 0.0555
40.50 Farnesol 222.0 0.0005 0.56 3.8228
41.30 Benzyl benzoate 212.0 0.0005 4.95 32.3404
43.66 Phenylethyl benzoate 226.0 0.0005 0.12 0.8151
44.27 Benzyl salicylate 228.0 0.0005 7.23 50.8120
flowers scent can be obtained by passing the collected sample through both polar
and a nonpolar GC columns. The eluent of each column is split, with one portion
going to a sniff-port and the other to an appropriate detector. Peaks that are not
detected at the sniff-port can be identified and the component can be eliminated
from the liquid formula.
Reconstitutions for commercial purposes also have to comply with regula-
tory agencies. This may require the perfumer to substitute a component of the
flowers scent with an olfactively similar aroma chemical and rebalance the for-
mula, if necessary.
B. Identification and Elucidation of New Molecules

As discussed, nature provided the source of inspiration for the synthesis of aroma
chemicals. This remains as true today as it was in the nineteenth century. One
of the reasons to explore the rain forests is biodiversity prospecting, which is the
search for novel chemicals from biological sources. Rain forests are home not
only to the majority of the worlds flowering plants but to the majority of insects
as well. This diversity of new flowers and pollinators increases the probability
of discovering new aroma chemicals.
The technique for identifying new aroma chemicals is discussed by Kaiser
(36). He utilized dynamic headspace using Porpak traps to collect unusual flower
scents. Unknown peaks present in the headspace and not identifiable by GC/MS
analysis were investigated. Priority was given to those peaks that had an olfactive
interest or a high odor intensity. The unknown aroma chemicals were isolated
by preparative GC and further identified by a combination of GC/MS, Nuclear
Magnetic Resonance and Nuclear Overhauser Effect experiments. This enabled
Kaiser to identify several new aroma chemicals. For example, in his study of
Houlletia odoratissima, a rare orchid native to South America, he found an un-
identified component with a molecular weight of 166. From spectral data he was
able to propose the structure of 7,7-dimethyl-4,5,6,7-tetrahydro-1 (3H)-isoben-
zopherone, which was subsequently confirmed by synthesis.
VI. CONCLUSIONS
Headspace techniques have enabled the fragrance chemist to continue the search
for new aroma chemicals and to recreate flower scents. For easily accessible
regions, the most appropriate methodology is dynamic headspace using a poly-
meric adsorbent such as Tenax, if thermal desorption will be used, or Porpak Q,
if solvent elution is to be used. This is preferable to SPME, which may require
the use of more than one fiber. SPME can, however, provide a useful complement
to dynamic headspace gas collection techniques for the capture of specific aroma
chemicals.
In remote regions, special techniques may be required to allow scientists

to identify when the flowers scent is at peak emission. Automatic, low-weight
collection devices have been designed to do this. Techniques have also been
developed to capture volatiles from previously inaccessible flowers, like those
blooming on the terminal branches on tall rain forest trees. Such a technique uses
the Zenith trap, which combines the benefits of SPME and dynamic headspace
in a practical system.
ACKNOWLEDGMENTS
The authors wish to extend special thanks to Amy Bekov for her constructive
comments on the manuscript, to Richard Sgaramella for his contributions to the
development and evaluation of the new techniques discussed, and to Katherine
Sznerch for carrying out the experiments. We would also like to thank Dragoco
for providing us with the drawings related to the work of Brunke et al.
REFERENCES
1. E. J. Parry, The Chemistry of Essential Oils, Scott Greenwood & Co., Aberdeen
University Press, UK, 1899.
2. D. H. Pybus, The history of aroma chemistry and perfumes, The Chemistry of Fra-
grance: Perfume Creativity from Brief to Market Place, ch. 2 (D. H. Pybus and
C. S. Sell, eds), Royal Society of Chemistry, Cambridge, UK, 1999.
3. B. D. Mookherjee and C. J. Mussinan, eds., 178th ACS meeting, 1979, Essential
Oils, Allured Publishing Company, Wheaton, Illinois, 1981.
4. R. Teranishi, R. E. Lundin, Analytical technique, Chemistry and Plants, ch. 7, ACS
Symposium Series 525, 1993.
5. R. Croteau and F. Karp, Origin of natural odorants, Perfumes: Art, Science and
Technology, ch. 4 (P. M. Muller and D. Lamparsky, eds.), Elsevier Applied Science,
London. UK, 1991.
6. R. Kaiser, The Scent of Orchids: Olfactory and Chemical Investigations, Elsevier,
Amsterdam, Netherlands, 1993.
7. T. P. Wampler. Analysis of food volatiles using headspace-gas chromatography
techniques, Techniques for Analyzing Food Aroma, ch. 2 (R. Marsili, ed), Marcel
Dekker, New York, 1997.
8. C. A. Dodson and H. G. Hills, Gas chromatography of orchid fragrances. Am. Orchid
Soc. Bull. 35, 720725 (1966).
9. E. J. Brunke, F. J. Hammerschmidt and G. Schmaus. The essential oil of Santolina
Chamaecyparissus, Dragoco Report 39, 151166 (1992).
10. D. Lamparski, Headspace techniques as a versatile complementary tool to increase
knowledge on constituents of domestic or exotic flowers and fruits, Essential Oils
and Aromatic Plants (B. A. Svendson and J. J. C. Scheffer, eds.), Martinus Nijoff,
Dordrecht, 1985, pp. 7992.
11. R. D. Barnes, L. M. Law and A. J. Macleod, Comparison of some porous polymers
as adsorbents for collection of odour samples and the application of the technique
to an environmental malodour, Analyst 106:412418 (1981).
12. R. Kaiser, Trapping investigation and reconstitution of flower scents, Perfumes: Art,
Science and Technology, ch. 7, (P. M. Muller and D. Lamparsky, eds.), Elsevier
Applied Science, London, UK, 1991.
13. E. J. Brunke, F. Hammerschmidt, G. Schmaus, Dragoco Report 1:331 (1992).
14. E. J. Brunke, F. Hammerschmidt, G. Schmaus, Dragoco Report 4:129145 (1996).
15. P. Berlardi and J. B. Pawlinszyn, The application of chemically modified fused silica
to capillary columns, Water Pollution Rev. J. Can. 24:179191, (1989).
16. J. Pawlinski, Solid Phase Microextraction: Theory and Practice, VCH, New York,
1997.
17. B. D. Mookherjee, S. M. Patel, R. W. Trenkle, and R. A. Wilson, Fragrance emission
from skin, Cosmetics Toiletries 113:5356 (1998).
18. X. Yang and T. Peppard, Solid-Phase microextraction for flavor analysis, J. Agric.
Food Chem. 42:19251930 (1994).
19. F. Pelusio, T. Nilsson, L. Montanarella, R. Tilio, B. Larsen, S. Facchetti and J. O.
Madsen, Headspace solid-phase microextraction analysis of volatile organic sulphur
compounds in black and white truffle aroma, J. Agric. Food Chem. 43, 21382143
(1995).
20. M. E. Miller and J. Stuart, Comparison of gas-sampled and SPME-sampled static
headspace for the determination of volatile flavor components, Anal. Chem. 71:23
27 (1999).
21. J. S. Elmore, M. A. Erbahdir, and D. S. Mottram, Comparison of dynamic headspace
concentration on Tenax with solid phase microextraction for the analysis of aroma
volatiles, J. Agric. Food Chem. 45:2632641 (1997).
22. A. Krumbein and D. Ulrich, Comparison of three sample preparation techniques for
the determination of fresh tomato aroma volatiles, Spec. Publ. R. Soc. Chem. 197:
289292 (1996).
23. T. McGee and K. L. Purzycki, Novel techniques for capturing scents in the Rainfor-
est Canopy, Perf. Flav. 24:15 (1999).
24. K. L. Purzycki and R. Sgaramella, Capturing volatiles emitted from flowers using
SPME, Eastern Analytical Symposium, Atlantic City, New Jersey, November 2000.
25. T McGee and K. L. Purzycki, Apparatus for obtaining or capturing odor chemicals,
European Patent Application EP 961108, 1998.
26. Sepelco, Sepelco Park, Bellefonte, PA.
27. B. D. Mookherjee, R. W. Trenkle, R. A. Wilson, M. Zampino, K. P. Sands, and
C. J. Mussinan, Fruits and flowers: live vs deadWhich do we want? Flavors and
Fragrance: A World Perspective. Proc. 10th Internat. Congress of Essential Oils,
Fragrance and Flavor, Washington 1986, (B. M. Lawrence, B. D. Mookerjee, and
B. J. Willis, eds.), Elsevier Science Publishers, Amsterdam, Netherlands, 1988,
pp. 415424.
28. R. Kaiser, Olfactory and Chemical Characteristics of Floral Scents, 196th National
ACS Meeting, Symposium on Progress in Essential Oil Research, Los Angeles,

1988.
29. L. Hornok, Effect of environmental factors on the production of essential oil plants,
Flavors and Fragrance: A World Perspective. Proc. 10th Internat. Congress of Essen-
tial Oils, Fragrance and Flavor, Washington 1986 (B. M. Lawrence, B. D. Mook-
erjee, and B. J. Willis, eds.), Elsevier Science Publishers, Amsterdam, Netherlands,
1988, pp. 129140.
30. R. Kaiser, Environmental scents at the Liguarian Coast, Perf. Flay. 22:718 (1997).
31. T. McGee and K. L. Purzycki, Givaudan-Roures ScentTrek searches for novel
scents, (S. Dorman, ed), Spray Technology Marketing, 3235 (1997).
32. M. W. Moffet, The High Frontier: Exploring the Tropical Rainforest Canopy, Har-
vard University Press, Cambridge, Mass, 1993.
33. Tedlar Bags, Alltech Associates, Inc., 2051 Waukegan Road, Deerfield, Illinois
60015.
34. N. Neuner-Jehle and F. Etzweiler, Perfumes Art Science & Technology, ch. 6, Elsev-
ier Science Publishers LTD, UK, 1991, p. 153.
35. Boiling and Vapor Pressure Calculator, Advanced Chemistry Development, 133
Richmond Street, West, Suite 605, Toronto, Ontario Canada, M5H 2L5.
36. R. Kaiser, New or uncommon Volatile compounds in floral scents, 13th Int. Congr.
Flavors and Fragrances and Essential Oils, October, 1519, Istanbul, Turkey, 1995.
10
SPME Applications in Consumer
Products
Richard Payne, Allen E. Puchalski, and John Labows
Colgate-Palmolive Company, Piscataway, New Jersey
I. INTRODUCTION
The odor character of a consumer product contributes to its uniqueness and can
be a signal of product efficacy. This odor impression is important during use of
the product, and for providing pleasant long-lasting aromas to clothes, in the case
of a fabric detergent or softener, or to the home with an all-purpose cleaner. It is
necessary to supplement subjective consumer or panel evaluations with analytical
methods to understand the performance of fragrance/flavor materials from
these products. In addition, any distortion of the intended odor can leave a nega-
tive impression with the consumer. Off-odors can occur because of quality issues
with the fragrance or its stability in the base and/or packaging. Malodors can
come from base materials or develop during processing. All these problems have
been addressed by various headspace techniques, including static and dynamic
headspace gas chromatography and gas chromatography/mass spectrometry
(GC/MS) and, more recently, solid-phase microextraction (SPME) headspace
coupled with chromatography. Imaginative sampling techniques are needed to
obtain fragrance/flavor profiles from the product in use where it is necessary to
sample from the mouth and skin for personal care applications and from fabric
and hard surfaces. In this chapter, applications will be described that show how
SPME analysis can be utilized to address both product and personal malodor and
fragrance/flavor performance issues.
SPME has become a valuable alternative to solvent extraction, purge-and-
trap (dynamic), and static headspace methods (14). This is true for the analysis
of flavors, fragrances, food aromas, and biological systems as is evidenced by
277
278 Payne et al.
numerous publications (511). The SPME technique provides more advantages

because it maintains the benefits of the traditional methods without the inherent
disadvantages (solvents, instrument cost, and sample preparation time). All sam-
ple matrices, liquids, solids, and gases, can be readily sampled. However, it
should be noted that inherently each headspace technique does give a unique
profile from the same sample based on equilibration and sampling times.
SPME does allow for the detection of less volatile materials, i.e., nitro- and poly-
cyclic musks, found in fragrances and not observed in static headspace methods
(12,13).
SPME methodology continues to improve with the development of differ-
ent fibers for specific applications (14). An alternative approach to increase sensi-
tivity, called Scentrek , has been described by Givaudan-Roure and involves the
bundling of short pieces of coated capillary GC columns to sample living plant
odors on location (15). Because-SPME injection requires only minor adjustment
to the GC injector, it is now possible to add automated sampling. Varian provides
two devices as accessories to their gas chromatographs (16). A heated liquid
sampling tray is convertible from use of a liquid syringe to sampling with an
SPME needle. A robotic sampling system, Combipal , allows larger sample num-
bers and heating and mixing. This makes SPME a valuable technique not only
for a research environment but also for routine monitoring. Sampling has also
been one of the key variables in the use of electronic sensors [nose] as odor
detectors. SPME has been incorporated via the Combipal as an odor sampling
method that provides a more representative headspace [vs static] for the odor
sensors for improved sample classification (17).
II. OFF-ODOR ANALYSIS

A. Fragrance Stability
The fragrance raw materials must be stable in aggressive consumer product bases,
i.e., those with extremes of pH or bleach formulas with peroxides or hypochlorite
as oxidants. As product bases are developed or undergo revisions, it is necessary
to test fragrance ingredients through an aging protocol. One such base change
to a high pH detergent formula resulted in the loss of many but not all ester
materials. Analysis of the ingredients in the headspace by SPME showed which
components were stable in the new base. Structurally hindered esters such as
verdox were stable under hydrolysis conditions relative to primary esters includ-
ing benzyl acetate, hexyl salicylate, and hexyl acetate and to other secondary
esters, cyclacet and vertenex. The resulting alcohols could be detected in the
product headspace. Figure 1 shows the reconstructed ion chromatogram (RIC)
from the GC/MS of the volatiles from two bases differing in pH and compares the
SPME Applications in Consumer Products 279
FIGURE 1 Fragrance profile comparison from detergent bases differing in pH.
change in intensity for the ingredients verdox (ortho-tert-butyl-cyclohexylacetate)

and vertenex (para-tert-butylcyclohexylacetate). Verdox is a hindered ester
and is more stable to hydrolysis in this product (Fig. 2). This emphasizes the need
to have simple, rapid methodology to monitor stability of fragrance ingredients as
new base compositions are developed.
B. Base Malodor
Malodors in raw materials contribute to unacceptable products as well as the
need to use higher levels of fragrance. One such example was the synthesis of
sodium lauryl ethoxy sulfate (SLES), a major surfactant used in personal care
products. This material is produced by ethoxylation of lauryl alcohol followed
by sulfonation. Often it is necessary to add a final bleaching step to minimize
280 Payne et al.
FIGURE 2 Chemical structural differences responsible for base stability/instability (ver-

dox, vertenex) and surfactant solubilization (citronellol, linalool).
the color of the SLES; however, in one case, this process introduced unacceptable
malodors to the SLES sample. SPME was used to sample the raw material before
and after bleach treatments. From the SPME/GC/MS the malodor was identified
as lauryl aldehyde produced by oxidation of residual lauryl alcohol. Because
SPME can be quickly adapted for injection into any GC, we were able to utilize
SPME with a GC adapted for olfactory studies. Sampling of the odorous material
using SPME/GC/olfaction confirmed the aldehyde as the unique odorant in the
SLES (Fig. 3).
FIGURE 3 Lauryl aldehyde identified as off-odor in SLES by SPME/GC/MS.

C. Packaging Malodor
Consumer complaints of off-odor in a major soap bar variant led to an exten-
sive investigation of the malodor cause. The odor appeared to come from both
packaging and wrapped soap bars. However subsequent olfactory analysis
indicated the strong off-odor was emanating from the packaging material. In
this case, the packaging was produced in one country but soap bars were
manufactured at another location. SPME was used to sample the headspace
above the wrapper system and soap bars. The wrapper system consists of a mid-
dle paper layer, an outer-oriented polypropylene layer with graphics and an
inner layer of microcrystalline wax. For analysis, two pieces of the soap wrap-
per or packaging raw material were placed in 4-ml vials and equilibrated
overnight at room temperature. They were sampled using a 65 m PDMS/
DVB (polydimethyl-siloxane/divinyl benzene) fiber exposed to each sample
for 15 minutes at room temperature. The fibers were desorbed at 225C for 5
minutes in the GC injector, and components were identified with an ion trap
detector.
Benzothiazole, a compound with a rubbery, burnt odor, was identified in
samples from both sources (Fig. 4). It was discovered that the paper mold-inhibi-
tor [2-(thiocyano-methylthio)-benzothiazole] was susceptible to thermal degrada-
FIGURE 4 Headspace SPME sampling of soap wrapper for the detection of benzothiazole,
a malodorant formed from the decomposition of a fungicide. A series of alkyl benzenes
was also observed between 3 and 7 minutes.
282 Payne et al.
tion and that during the application of a hot microcrystalline wax to the wrapper
the mold-inhibitor decomposed to benzothiazole and several other materials. The
benzothiazole then migrated from the soap wrapper to the soap bar resulting in
an altered odor profile. SPME analysis facilitated the identification of the source
as the chemical decomposition of mold inhibitor and the point in the process
where this occurs and allowed for the duplication of the process in the laboratory
(18).
D. Packaging Figurines
There is a continual need to update product formulas and packaging. One innova-
tion in packaging was the unique inserts to give three-dimensional scenes in Soft-
soap liquid handsoap. Another packaging effort involved the use of small premi-
ums (or figurines) within the product. Figurines were manufactured from PVC
containing a mixture of octyl adipate and benzoic acid glycerol esters as the
plasticizer. The figurines were painted with pigments dispersed in cyclohexanone
and acetone. After the figurine was aged in product base, a marked reduction in
viscosity and sharp off-odors were observed.
SPME analysis was performed on the painted and unpainted figurines using
a 100 m PDMS fiber exposed to the headspace of the sample for 15 minutes.
In order to determine if residual solvents were leaching into the product, a solu-
tion of base was prepared using 25% NaCl as the diluent. Five milliliters of
sample were transferred to a 10-mL vial and thermostatted for 1 hour at 30C
and sampled for 15 minutes. Chemicals typical of PVC were identified in both
painted and unpainted figurines as well as in the product base, and residual paint
volatiles could be detected from painted figurines. These included toluene, cyclo-
hexanone, 2-ethyl hexanol, 2-butoxyethanol, and a series of alkyl benzene. These
findings are consistent with previous studies conducted on PVC for volatile or-
ganic compounds (1921). Figure 5 shows the headspace SPME profiles of
painted and unpainted PVC and base product solutions.
E. Volatile Exposure
Formaldehyde is an airborne contaminant from various environmental sources.
Acceptable concentration limits of formaldehyde in ambient air are in the range
of 20100 ppbv. It is not possible to sample formaldehyde directly by SPME.
However, a unique SPME headspace sampling approach has been developed
to quantitatively sample formaldehyde (22). This method involves first exposing
the fiber to o-pentafluorobenzyl-hydroxylamine hydrochloride and then to the
product headspace. The formaldehyde is derivatized directly on the fiber and
after desorption in the GC is quantitated as the derivative. This methodology is
FIGURE 5 Selected ion chromatograms for headspace SPME sampling of liquid hand
soap base aged with a painted figurine. Top (A) chromatogram shows cyclohexanone as
major component. A series of alkyl benzenes were also identified (B) in the product base.
The numbered peaks are toluene (1), ethylbenzene (2), m-xylene (3), p-xylene (4). All
compounds were determined to be less than 20 ppm in the liquid hand soap.
applicable to consumer products because formalin is often the preservative of

choice and has been applied to shampoo and particleboard (22). An example of
the application of this approach involves the preservative in a dishwashing liquid.
The ion chromatogram in Fig. 6 shows the reagent itself and formaldehyde as
the oxime derivative from the headspace of a formalin-containing dishwashing
liquid.
284 Payne et al.
FIGURE 6 Formaldehyde detected as oxime derivative in headspace of a dish liquid.
III. ORAL MALODOR
Various methods have been used to investigate oral malodors including direct
oral sampling for volatile sulfur materials and dynamic headspace analysis of
incubated saliva. These approaches have been used in clinical evaluation of mate-
rials to ameliorate oral malodors as well as identifying all materials that contribute
to the malodor or are indicators of oral disease (2327). SPME analysis has been
used to sample human breath (28). In this study, systemic air from diabetics was
analyzed for ethanol, acetone, and isoprene.
We have recently reported the use of SPME to study salivary volatile sulfur
compounds (29). SPME analysis of incubated whole saliva in culture medium
(sampled for 5 minutes using a 75 m CAR/PDMS fiber; equilibration time of
15 minutes) led to the detection of a wide range of compounds that are known for
TABLE 1 Compounds Found in Salivary Headspace by Static HS-SPME a
Ret. time
Peak no. Compounds (min) ID b
1 Methyl Mercaptan 1.55 rt, ms
2 Dimethyl sulfide 1.60 rt, ms
3 Acetone 1.80 rt, ms
4 1-(Methylthio)-propane 2.25 T, ms
5 Z-1-(Methylthio)-1-propene* 2.77 T, ms
6 2-Methyl butanoic acid methyl ester* 2.88 T, ms
7 E-1-(Methylthio)-1-propene* 2.97 T, ms
8 Toluene c 3.20 rt, ms
9 S-Methyl ethanethioate 3.27 T, ms
10 Dimethyl disulfide 3.62 rt, ms
11 4-Methyl pentanoic acid methyl ester* 4.27 T, ms
12 Limonene c 4.98 rt, ms
13 S-Methyl 3-methylbutanethioate* 5.27 T, ms
14 S-Methyl pentanethioate* 5.38 T, ms
15 Thiocyanic acid methyl ester* 6.08 T, ms
16 2,5-Dimethyl pyrazine c 8.37 rt, ms
17 2,2-Bis(methylthio)propane* 7.40 T, ms
18 Dimethyl trisulfide 7.67 rt, ms
19 Alkyl benzene c 8.13 T, ms
20 Dimethyl sulfoxide* 10.05 rt, ms
21 Methyl(methylthio)methyl disulfide 11.42 T, ms
22 Phenol 15.37 rt, ms
23 Indole 23.43 rt, ms
a
Samples equilibrated for 15 minutes and sampled for 5 minutes.
b
Identification was by GC retention times (rt) and mass spectrometry (ms) of authentic compounds.
Tentative (T) identification by mass spectrometry only when authentic compound was not available.
Peaks 22 and 23 not shown in Figure 7.
c
Probable exogenous sources.
* Not previously identified in saliva.
their malodor characteristics (Table 1). These include expected sulfur metabolites
based on methyl mercaptan, such as dimethylsulfide (DMS), dimethyl disulfide
(DMDS), dimethyl trisulfide (DMTS) and methylthioesters, and propionic and
acetic acids, phenol (PH) and indole (IN). Normalization of each peak of the
chromatographic profile of incubated saliva to the total peak area showed that
approximately 90% of the headspace consisted of sulfur metabolites and phenol
and indole (Fig. 7). In addition there were a large number of minor components,
described in Table 1. Hydrogen sulfide, which is a key oral malodorant, was not
observed in our chromatographic profiles using porous Carboxen as the adsor-
286 Payne et al.
FIGURE 7 SPME/GC/MS of headspace of incubated saliva. Numbers refer to compounds

identified in Table 1. Peaks labeled F were identified as contaminants from both fiber and
vial septa.
bent (fiber coating). We believe this is a failure to absorb on the fiber and as a
result hydrogen sulfide was not effectively trapped.
This study was notable in the use of SPME and its ability to effectively
sample and identify three S-methyl thioesters, two of which have not been re-
ported (S-methyl 3-butanethioate and S-methyl pentanethioate). Identification
was based on comparison with library spectra and/or through interpretation of
the collected mass spectra. Although it has been established that HS, MM, and
DMS are the major constituents of oral malodor, other minor components such
as the S-methyl thioesters that have exceptionally low odor thresholds may con-
tribute to the oral malodor. The presence of S-methyl ethanethioate in saliva
and tongue scrapings has been reported (26). Interestingly, methyl-(methylthio)-
methyl disulfide (peak 21 in Table 1) was identified in saliva, whereas it had
only been observed previously in tongue scrapings after incubation in casein,
suggesting incubation media may play a role in the formation of VSC. Peaks
labeled with an asterisk in Table 1 are of particular interest because they have
not been previously reported in saliva headspace.
IV. FRAGRANCE PERFORMANCE

A. Fragrance Solubilization
Surfactants, both anionic and nonionic, in cleaning products act to solubilize and
remove the dirt/grease from substrates such as hard surfaces and fabric. In solu-
tion the surfactants form micelle structures that also solubilize product ingredients
such as the fragrance raw materials. This can significantly alter the fragrance
character of the perfume oil; thus, fragrances are created for specific bases. Head-
space techniques have been used to measure the availability of individual ingredi-
ents for perception and to show how in-use dilution of the base changes the
headspace profile (30). Static headspace analysis limits the range of fragrance
materials that can be detected. SPME analysis provides a profile of most ingredi-
ents including those of low volatility with sensitivity to follow those ingredients
over a wide range of product dilution.
Figure 8 compares the SPME headspace/GC/MS profile of an all-purpose
cleaner (APC) formula neat and as a 256-fold dilution in water. These time points
correspond to point of purchase and fragrance bloom. Some ingredients main-
tain their headspace concentration whereas others decrease significantly at this dilu-
tion. Figure 9 charts the change in headspace level of five ingredients throughout
the dilution sequence, i.e., series of half-fold dilutions in water. Those chemicals,
FIGURE 8 SPME headspace of an APC product both neat and as a 256-fold dilution.
288 Payne et al.
FIGURE 9 Change in headspace profile over a half-fold dilution sequence for several
fragrance components. Components that are more soluble in the surfactant persist longer
in the headspace and contribute more to the odor in the diluted product.
such as citronellol, that are more soluble in the surfactant micelle (i.e., have struc-
tural similarity to surfactants) are able to maintain their headspace level over many
dilutions. Others with floral character that are either more water soluble, such as
benzyl acetate, or less soluble in the micelle due to structural constraints (linalool
is a tertiary alcohol; citronellol is a primary alcohol) decrease significantly in the
headspace (Fig. 2). This approach gives a picture of the possible fragrance character
change on usage of an APC represented as the fragrance bloom.
B. Fragrance SubstantivityCloth/Tile
Many products advertise long-lasting fresh clean odors on fabric or in your home.
These claims are often subjective panel judgments but they can be supported by
analytical measurements from the respective substrates. Two applications of
SPME have involved developing sampling methods to measure substantivity
from cloth and tile.
Dynamic headspace has been used to measure substantivity of fragrance
ingredients on cloth after rinsing with a fabric softener (31). These measurements
were made at up to 10 days and compared with odor assessments. The factors
influencing the substantivity include vapor pressure, water solubility, structure
of matrix, and odor threshold for sensory analysis. In experiments in our lab,
SPME was used to measure effects of softener formula on fragrance delivery
and substantivity with cloth. Cloth swatches were washed in unfragranced deter-
gent and then rinsed in a fragranced fabric softener. After air drying the cloth
was placed in headspace vials and equilibrated for 30 minutes at 55C and sam-
pled with 100 m PDMS fiber for 10 minutes (32). Figure 10 shows the profile at
10 days post wash. The fabric retains the fragrance character and the substantive
components are clearly detectable off the fabric. This approach allows the com-
parison of formulation changes and the substantiation of product claims.
FIGURE 10 GC/MS fragrance profile from cloth sampled by SPME 10 days following
a rinse with fabric softener. Compounds include lilial (L), hexyl cinnamic aldehyde
(HCA), and galaxolide (G).
Because APCs are often formulated for use as direct-spray application,

SPME can be used to analytically compare the initial impact with the residual
fragrance remaining after dry-out. A 6-inch square Formica tile was sprayed with
a modified paint sprayer to provide a uniform, reproducible coating of cleaner.
A small resin kettle head, with outlet connectors fitted with septa to allow for
SPME sampling, is then placed in the center of the tile. An O-ring provides a
seal between the tile and the kettle head. The headspace is allowed to equilibrate
for 30 minutes and the SPME fiber (100 m PDMS) is then exposed to the head-
space for 10 minutes. The resin kettle head is removed and the tile is left open
to the atmosphere until the next equilibration period. Figure 11 shows an initial
sample with high levels of the more volatile fragrance materials and the same
tile after 3.5 hours. The 3.5 hour sample still has measurable amounts of several
fragrance materials, demonstrating the benefit of substantive materials to a long-
lasting impression of performance. These results can be compared with sampling
from neat solutions (see Sec. IV.A) that measures initial impact and in-use dilu-
tion for odor impression.
C. Fragrance SubstantivitySkin
The efficacy of many personal products relies on their ability to deliver fragrances
to the skin. In the case of deodorants, claims of 24-hour odor protection are based
on the substantive nature of the fragrance ingredients in combination with the
base to mask human malodor. With shampoos, soaps, and fine fragrances, it is
290 Payne et al.
FIGURE 11 Fragrance profile sampled at 0.5 and 3.5 hours from tile following application
of an APC.
the impression of the fragrance itself that is delivered to the skin. Initially body
tubes and telephone booth type sampling chambers were used with purge-
and-trap techniques to sample total body malodors. More recently dynamic sam-
pling has been used to sample fragrances or products applied to the skin (33
34). This approach immobilizes an individuals arm in a glass sleeve and allows
sampling from skin at different times following sample application or measures
diffusion by sampling at different distances from point of application. More re-
cently SPME has been applied by IFF to sample the fragrance aura from living
flowers and from skin following application of fine fragrance (35).
In the latter technique, a glass-sampling bulb is placed over the skin of the
forearm with an inlet for the SPME sampling needle. The fiber tip is placed about
1 cm above the skin for 3060 minutes and then the sample is analyzed by GC/
MS. The components observed were the top notes as expected but also some of
the more diffusive components that were not observed from the oil itself. Some
sulfur-containing materials possessed high diffusivity, low thresholds, and high
vapor pressures and thus had a high impact. Other high-molecular-weight, low-
volatile materials also had high diffusivity and composed a large part of the
aura of the skin. This approach suggests that SPME could simplify the time
and apparatus necessary to sample odorants from the skin. Dynamic SPME as
described below may also be applicable in this case.
V. FLAVOR SUBSTANTIVITY
The effects of dentifrice components on flavor release in vitro using static head-
space had previously been studied (3637). Recently consumer emphasis on
long-lasting flavor has led to extensive work in understanding the release of flavor
in the oral cavity (38). Atmospheric pressure chemical ionization techniques have
FIGURE 12 Dynamic headspace SPME sampling apparatus.

292 Payne et al.
been adapted to sample mouth and/or nose headspace in real time (3940). This
results in the ability to profile flavor materials either retronasally or directly from
the mouth as they are released from the matrix of a food or oral product. Though
this is an ideal analytical method for measuring flavor release, it requires a consid-
erable initial equipment investment (APCI interface and special mass spectrome-
ter) compared with that for SPME.
SPME was utilized to study the substantivity of individual flavor compo-
nents in mouth following brushing with a dentifrice. Both direct and dynamic
SPME sampling allowed the release profile of the major flavorants to be studied.
Using a simple glass mixing/sampling chamber, it was possible to sample a turbu-
lent stream of air from the oral cavity (Fig. 12). Although reproducibility between
subjects was approximately 20%, the distribution of flavor chemicals (area %)
was consistent. Figure 13 shows a GC profile of flavorants release from the mouth
FIGURE 13 SPME profile of flavor volatiles sampled from the mouth at 5 and 30 minutes
following brushing with a dentifrice. Peaks are identified as Menthone (1), Isomenthone
(2), Menthyl acetate (3), Isomenthol (4), Menthol (5), Methyl salicylate (6), and Anethole
(7).
1 minute after brushing with dentifrice. A detailed investigation of sampling pa-

rameters from the mouth is necessary to understand the differences between sub-
jects. However, using this simple method it is possible to evaluate the effects of
dentifrice formulation on flavor longevity in the oral cavity.
VI. SUMMARY
As shown through the above examples, SPME has been an extremely useful
method for sampling volatiles from a variety of matrices. The data can be used to
profile the performance of flavor and fragrance materials when applied to various
substrates. Coupled with mass spectrometry, SPME allows identification of fra-
grance materials among peaks from product base and the ability, with reasonable
sampling times, to detect the less volatile fragrance materials such as musks,
coumarin, and salicylate derivatives. SPME also provides a rapid method to sam-
ple and potentially identify chemicals that contribute to product off-odors.
REFERENCES
1. R. Marsili, SPME-MS-MVA as an electronic nose for the study of off-flavors in

milk, J. Agric. Food Chem. 47:648654 (1999).
2. R. Marsili, Comparison of solid-phase microextraction and dynamic headspace
methods for the gas chromatographic-mass spectrometric analysis of light-induced
lipid oxidation products in milk, J. Chrom. Sci. 37:1723 (1999).
3. S. Scypinski and A. Smith, Pharmaceutical applications, Solid Phase Microextrac-
tion: A Practical Guide (S. Scheppers-Wercinski, ed.), Marcel Dekker, New York,
1999, pp. 111129.
4. J. R. Almirall and K. Furton, Forensic and toxicology applications, ibid., pp. 203
216.
5. J. Song, L. Fan, and R. Beaudry, Application of solid phase microextraction and
gas chromatography/time-of-flight mass spectrometry for rapid analysis of flavor
volatiles in tomato and strawberry fruits, J. Agric. Food Chem. 46:37213716
(1998).
6. A. Harmon, Solid-phase microextraction for the analysis of flavors, Techniques for
Analyzing Food Aroma (R. Marsili, ed.) Marcel Dekker, New York, 1997, pp. 81
112.
7. K. Krumbein and D. Ulrich, Comparison of three sample preparation techniques for
determination of fresh tomato aroma volatiles, Flavor Science: Recent Developments
(A. Taylor and D. Mottram, eds.), The Royal Society of Chemistry, Cambridge, UK,
1996, pp. 289292.
8. R. Galipo, A. Canhoto, M. Walla, and S. Morgan, Analysis of volatile fragrance
and flavor compounds by headspace solid phase microextraction and GC-MS, J.
Chem. Educ. 76:245248 (1999).
294 Payne et al.
9. D. Roberts and R. Pollien, Analysis of aroma release during microwave heating, J.

Agric. Food Chem. 45:43884392 (1997).
10. T. Braggins, C. Grimm, and F. Visser, Analysis of food volatiles using SPME, Appli-
cations of Solid Phase Microextraction (J. Pawliszyn, ed.), The Royal Society of
Chemistry: Cambridge, UK, 1999, pp. 407422.
11. G. Moneti, G. Pieraccini, M. Sledge, and S. Turillazzi, SPME in the study of chemi-
cal communication in social wasps, ibid., pp. 448457.
12. J. Reilly, R. Payne, D. Kutylowski, and J. Labows, Polycyclic musks in consumer
detergent products using headspace SPME/GC/MS, 37th Eastern Analytical Sympo-
sium, November 1998.
13. C. Struppe, B. Schafer, and W. Engewald, Nitro musks in cosmetic productsdeter-
mination by headspace solid-phase microextraction and gas chromatography with
atomic-emission detection, Chromatography 45:138144 (1997).
14. R. Shirey and V. Mani, New trends and applications using solid phase microextrac-
tion, Technical Report 1997; Supelco, Bellefonte, PA.
15. T. McGee and K. Purzycki, Novel Techniques for Capturing Scents in the Rainforest
Canopy: the Final Frontier, Perfumer & Flavorist 24:18 (1999).
16. Varian, Inc., Walnut Creek, CA.
17. Alpha MOS, Toulesse, France.
18. R. Payne, Applications of solid phase microextraction in personal care products,
38th Eastern Analytical Symposium, November 1999.
19. D. Apostolopoulos, Identification of the source of an off-odor in premiums intended
for use with dry mix beverages, Food Flavors: Formation, Analysis and Packaging
Influences (E. Contis, C. T. Ho, C. Mussinan, T. Parliment, F. Shahidi, and A.
Spanier, eds.), Elsevier, Amsterdam, 1998, pp. 753757.
20. L. Gramiccioni, S. Di Marzio, P. Di Prospero, and C. Arena, A rapid headspace gas
chromatographic method for the determination of residual cyclohexanone in PVC
biomedical devices, Rass. Chim. 37:259261 (1985).
21. K. Slebioda and R. Wiglusz, Multiple headspace extraction-gas chromatography
analysis of residual solvents in PVC floor coverings. Chem. Anal. (Warsaw) 42:
373376 (1997).
22. P. A. Martos and J. Pawliszyn, Sampling and determination of formaldehyde using
solid-phase microextraction with on-fiber derivatization, Anal. Chem. 70:2311
2320 (1998).
23. J. Tonzetich, J. Coil, and W. Ng, Gas chromatographic method for trapping and
detection of volatile organic compounds from human mouth air, J. Clin. Dent. 2:
7982 (1991).
24. J. Kostelc, G. Preti, P. Zelson, N. Stoller, and J. Tonzetich, Salivary volatiles as
indicators of periodontitis, J Periodontal Res., 15:185192 (1980).
25. G. Preti, L. Clark, B. Cowart, R. Feldman, D. Lowry, E. Weber, and I. Young, Non-
oral etiologies of oral malodor and altered chemosensation, J. Periodontol. 63:790
796 (1992).
26. Y. Ghoos, D. Claus, B. Geypens, M. Hiele, B. Maes, and P. Rutgeerts, Screening
method for the determination of volatiles in biomedical samples by means of an off-
line closed-loop trapping system and high-resolution gas chromatography-ion trap
detection, J. Chromatogr. A 665:333345 (1994).
27. D. Claus, B. Geypens, Y. Ghoos, and P. Rutgeerts, Oral malodor assessed by closed-
loop gas chromatography and ion-trap technology, J. High Resol. Chromatogr. 20:
9498 (1997).
28. C. Grote and J. Pawliszyn, Solid-phase microextraction for the analysis of human
breath, Anal. Chem. 69:587596 (1997).
29. R. Payne, S. Liu, and J. Labows, Released oral malodors measured by SPME/GC/
MS, Flavor Release (D. Roberts and A. Taylor, eds.), American Chemical Society,
Washington, DC, 2000, pp. 7386.
30. J. Brahms, J. Labows, and R. Cagan, Solubilization of fragrances by surfactants,
Surfactants in Cosmetics (L. Rhein and M. Rieger, eds.), 2nd ed., Marcel Dekker,
New York 1997, pp. 605619.
31. X. Ormancey, S. Puech, and D. Coutiere, Substantivity of fragrances on cloth, Per-
fumer & Flavorist, 25:2429 (2000).
32. D. Smith and M. Heibel, Colgate-Palmolive, personal communication.
33. A. Baydar, A. Charles, J. Decazes, T. McGee, and K. Purzycki, Behavior of fra-
grances on skin, Cosm. & Toil. 111:4957 (1996).
34. C. Vuilleumier, I. Flament, and P. Sauvegrain, Headspace measurements of evapora-
tion rates of perfumes applied to skin, Perfumer & Flavorist 20:19 (1995).
35. B. Mookherjee, S. Patel, R. Trenkle, and R. Wilson, Fragrance emission from the
skin, Cosm. Toil. 113:5360 (1998).
36. J. Brahms, J. Masters, J. Labows, and M. Prencipe, Investigation of the interaction
between dentifrice flavor and product base by principal component analysis of head-
space-GC data, FlavorFood Interactions (J. Leland and R. McGorrin, eds.), ACS
Symposium Series 633, 1996, pp. 188200.
37. J. Brahms, J. Labows, and R. Cagan, Solubilization of Flavors, Contributions of
Low and Non-Volatile Materials to the Flavor of Foods (W. Pickenhagen, C-T. Ho
and A. Spanier, eds.), Allured Publishing, Carol Stream, IL 1996, pp. 125135.
38. D. Roberts and A. Taylor, Flavor release: a rationale for its study, Flavor Release
(D. Roberts and A. Taylor, eds.), American Chemical Society, Washington, DC,
2000, pp. 16.
39. A. Taylor and R. Linforth, Techniques for measuring volatile release in vivo during
consumption of food, ibid., pp. 821.
40. R. Linforth, K. Ingham, and A. Taylor, Time course profiling of volatile release
from foods during the eating process, Flavor Science: Recent Developments (A.
Taylor and D. Mottram, eds.), The Royal Society of Chemistry, Cambridge, UK,
1996, pp. 361368.
11
Gas ChromatographyOlfactometry
in Food Aroma Analysis
Imre Blank
Nestle Research Center, Lausanne, Switzerland
I. INTRODUCTION
Progress in instrumental analysis has led to long lists of volatiles (1). Unfortu-
nately, the sensory relevance of these volatile compounds has not been as exten-
sively evaluated, although the use of the human nose as a sensitive detector in
gas chromatography (GC) was proposed by Fuller and coworkers as early as
1964 (2). In the meantime, much has been published on food aroma, often without
identifying the impact compounds. Therefore, one of the major problems in
aroma research is to select those compounds that significantly contribute to the
aroma of a food.
Flavor is usually divided into the subsets of taste and smell, which are
perceived in the mouth and the nose, respectively. However, flavor is fre-
quently used in publications exclusively dealing with volatiles. The terms
aroma and odor are not well defined and are often used as synonyms.
Odor is best reserved for the smell of food before it is put into the mouth (nasal
perception), and aroma for the retronasal smell of food in the mouth.
In general, the aroma of a food consists of many volatile compounds, only
a few of which are sensorially relevant. A first essential step in aroma analysis
is the distinction of the more potent odorants from volatiles having low or no
aroma activity. In 1963, Rothe and Thomas calculated the ratio of the concentra-
tion of an odorant to its odor threshold and denoted it aroma value (3). This
approach was the first attempt to estimate the sensory contribution of single odor-
ants to the overall aroma of a food. Since that time, similar methods have been
developed: odor unit (4) based on nasal odor thresholds, flavor unit (5) using
297
298 Blank
retronasal odor thresholds, and odor activity value (OAV) (6). However, this
concept requires identification and quantification of a great number of volatile
compounds and determination of their threshold values, which is time-consum-
ing. Furthermore, there is no guarantee that all of the important odorants were
considered, unless a screening step for the most important aroma compounds
was used.
GC in combination with olfactometric techniques (GC-O) is a valuable
method for the selection of aroma-active components from a complex mixture
(7). Experiments based on human subjects sniffing GC effluents are described
as GC-O. This technique helps to detect potent odorants, without knowing their
chemical structures, which might be overlooked by the OAV concept (ratio of
concentration to threshold) if the sensory aspect is not considered from the very
beginning of the analysis. Experience shows that many key aroma compounds
occur at very low concentrations; their sensory relevance is due to low odor
thresholds. Thus, the peak profile obtained by GC does not necessarily reflect
the aroma profile of the food.
The purpose of this contribution is to discuss recent developments in food
aroma analysis from the chemists point of view. It will particularly focus on
qualitative aroma composition obtained by GC-O. Potential and limitations of the
GC-O approach will be discussed and comments made to allow a more realistic
interpretation of data. This overview is addressed to flavor scientists from both
industry and academia.
II. GAS CHROMATOGRAPHYOLFACTOMETRY

TECHNIQUES
In general, it is very difficult to judge the sensory relevance of volatiles from a

single GC-O run. Several techniques have been developed to objectify GC-O
data and to estimate the sensory contribution of single aroma components. This
issue seems to be of great concern, as a considerable part of the 7th Weurman
Symposium was dedicated to this topic (8). Dilution techniques and time-inten-
sity measurements are the two main GC-O methods.
A. Time-Intensity Measurements
McDaniel et al. (9) have developed the technique Osme, measuring the perceived
odor intensity of a compound in the GC effluent. The subject rates the aroma
intensity by using a computerized 16-point scale time-intensity device and indi-
cates the corresponding aroma characteristics. This technique provides an FID-
style aromagram called an osmegram (Fig. 1). Ideally, it requires only one injec-
tion when working with well-trained assessors. Human subjects were found to
GC-O in Food Aroma Analysis 299
FIGURE 1 Osmegram with odor duration time (d) and maximum odor intensity (I max).
(From Ref. 10.)
be reliable instruments for reporting odor intensity changes in response to

changes in odorant concentration (10). Similar methods based on olfactive inten-
sity measurements have recently been reported (11).
B. Dilution Techniques
Two techniques based on dilution have been developed: CharmAnalysis by Acree
and coworkers (6,12,13) and aroma extract dilution analysis (AEDA) by Grosch
and his group (7,14,15). Both evaluate the odor activity of individual compounds
by sniffing the GC effluent of a series of dilutions of the original aroma extract.
Both methods are based on the odor-detection threshold. The dilution value ob-
tained for each compound is proportional to its OAV in air, i.e., its concentration.
Several injections are required to reach a dilution of the aroma extract in which
odorous regions are no longer detected.
In CharmAnalysis, the dilutions are presented in randomized order to avoid
bias introduced by knowledge of the samples. The assessor detects the beginning
and the end of each aroma perception (duration of the smell) and notes the sensory
attributes (Fig. 2). The dilution value is measured over the entire time of the
eluting peak. From these data, the computerized system constructs chromato-
graphic peaks where the peak areas are proportional to the amount of the odorant
in the extract. The Charm value is calculated according to the formula c d n1,
where n is the number of coincident responses and d the dilution value. The result
300 Blank
FIGURE 2 Schematic procedure for gas chromatographyolfactometry using Charm-

Analysis. (From Ref. 12.)
of the CharmAnalysis is displayed in a Charm chromatogram. As shown in Fig.

3, the sensory relevant volatile glucose-proline reaction products were detected
by CharmAnalysis and then, based on these results, identified as detailed in Fig.
4. The major peak found in the sample, 5-acetyl-2,3-1H-pyrrolizine, was almost
odorless (16).
FIGURE 3 Charm chromatograms of the volatile 200C glucose-proline reaction products

on OV 101 and Carbowax 20M columns. (From Ref. 16.)
FIGURE 4 Major odor-potent compounds in the glucose-proline reaction detected by

CharmAnalysis: diacetyl (no. 1), 2-acetyl-1-pyrroline (no. 2), 2-acetyl-1,4,5,6-tetrahydro-
pyridine (no. 3), 2-acetyl-3,4,5,6-tetrahydropyridine (no. 5), and furaneol (no. 6). Com-
pound numbers refer to Figure 3.
In AEDA, the assessor indicates whether or not an aroma can be perceived

and notes the sensory descriptor. The result is expressed as the flavor dilution
(FD) factor that corresponds to the maximum dilution value detected, i.e., the
peak height obtained in CharmAnalysis. The FD factor is a relative measure and
represents the odor threshold of the compound at a given concentration. The data
are presented in an FD chromatogram (Fig. 5) indicating the retention indices
(x-axis) and FD factors in a logarithmic scale (y-axis). AEDA has been proposed
as a screening method for potent odorants as the results are not corrected for
losses during isolation (7).
FIGURE 5 FD chromatogram of an aroma extract obtained from roast and ground Arabica
coffee. (From Ref. 17.)
302 Blank
The results obtained for freshly roasted Arabica coffee are illustrated in
Fig. 5 (17). From more than 1000 volatiles detected in the original aroma extract
by FID, only about 60 odor-active regions were selected by GC-O. AEDA re-
vealed 38 odorants with FD factors of 16 or higher. Odorants 5, 14, 19, 26, 30,
and 32 have been newly identified in coffee aroma. Their identification stemmed
from the high FD factors. They would most likely have been overlooked without
using GC-O as a screening method for odor-active compounds. Odorants with
FD factors of 128 or higher are shown in Fig. 6.
C. Static Headspace GC-O

The GC-O techniques described above mainly deal with aroma extracts (liquids)
isolated from the food. Recently, Guth and Grosch reported a new concept in
aroma research using static headspace in combination with GC-O (18). The
FIGURE 6 Chemical structures of some aroma impact compounds (FD 128) found
in an aroma extract of roast and ground Arabica coffee: 2-methyl-3-furanthiol (no. 5),
2-furfurylthiol (no. 6), methional (no. 8), 3-mercapto-3-methylbutyl formate (no. 14),
3-isopropyl-2-methoxypyrazine (no. 15), 2-ethyl-3,5-dimethylpyrazine (no. 17), 2-
ethenyl-3,5-dimethyl-pyrazine (no. 19), 2,3-diethyl-5-methylpyrazine (no. 21), 3-isobutyl-
2-methoxypyrazine (no. 25), 2-ethenyl-3-ethyl-5-metylpyrazine (no. 26), sotolon (no. 30),
4-ethylguaiacol (no. 31), abhexon (no. 32), 4-vinylguaiacol (no. 34), and (E)--damas-
cenone (no. 35). The numbering corresponds to that in Figure 5.
equipment is composed of a purge-and-trap system for introducing various vol-

umes of gaseous samples without artefact formation, a suitable capillary column,
and an effluent splitter to simultaneously perform GC-O and detection by FID
or MS (Fig. 7). A defined volume of the headspace is injected into a precooled
trap to focus the volatiles. After flushing the air present in the gas volume, GC
separation is started by raising the oven temperature. Dilution steps are made by
injecting decreasing headspace volumes to evaluate the relative odor potencies.
The problem of identification is solved by using the same analytical conditions
(capillary, temperature program) as for AEDA, so that identification can be per-
formed on the basis of odor qualities and RI values (18).
The sensory relevance of individual odorants can be estimated by injecting
various headspace volumes. This is equivalent to AEDA of liquid samples. In
contrast to AEDA, where aroma compounds are separated from the food matrix,
static headspace GC-O provides data about the aroma above the food. This tech-
nique is suitable for studying the effect of the food matrix on the aroma profile.
Therefore, AEDA and static headspace GC-O result in complementary data.
The analysis of coffee aroma is an excellent example of the potential of
static headspace GC-O (19). Compared to AEDA (17), compounds 14 and 7
were additionally detected (Table 1). The sensory contribution of odorants was
different from that obtained by AEDA. In general, very volatile compounds were
FIGURE 7 Schematic presentation of the static headspace GC-O technique. (Adapted

from Guth and Grosch, Annual Report of the Deutsche Forschungsanstalt fur Lebensmit-
telchemie, 1993, p. 27.)
304 Blank
TABLE 1 Aroma Impact Odorants of Roast and Ground Arabica and Robusta Coffee
Detected by Static Headspace GC-O a and Expressed as FD Factors b
FD factor
Aroma quality
No. Compound (on GC-O) Arabica Robusta
1 Acetaldehyde Fruity, pungent 25 25
2 Methanethiol Cabbagelike, sulfury 5 12.5
3 Propanal Fruity 5 5
4 Methylpropanal Fruity, malty 5 5
5 Diacetyl Buttery 62.5 125
6 3-Methylbutanal Malty 12.5 25
7 2-Methylbutanal Malty 5 12.5
8 2,3-Pentanedione Buttery 125 125
9 3-Methyl-2-butene-1-thiol Sulfury, Allium-like, foxy c 62.5 62.5
a
Modified from Ref. 19.
b
The headspace volumes of 25, 5, 2, 1, 0.4, and 0.2 ml correspond to the FD factors of 1, 5, 12.5,
25, 62.5, and 125, respectively.
c
Aroma qualities depend on concentration.
underestimated by AEDA, most likely due to losses during sample preparation.

2,3-Pentanedione (no. 8), diacetyl (no. 5), 3-methyl-2-butene-1-thiol (no. 9), acet-
aldehyde (no. 1), and 3-methylbutanal (no. 6) are key odorants of Arabica coffee
(Fig. 8). Methanethiol (no. 2) and 2-methylbutanal (no. 7) contribute more sig-
nificantly to the aroma of Robusta coffee.
FIGURE 8 Chemical structures of some aroma impact compounds (FD 12.5) found in
the headspace of roast and ground Arabica and Robusta coffee: acetaldehyde (no. 1),
methanethiol (no. 2), diacetyl (no. 5), 3-methylbutanal (no. 6), 2-methylbutanal (no. 7),
2,3-pentanedione (no. 8), and 3-methyl-2-butene-1-thiol (no. 9). The numbers correspond
to those in Table 1.
III. POTENTIAL OF THE GC-O APPROACH

A. Screening and Identification of Potent Odorants
Detection of odorous regions in a gas chromatogram is the first useful information
that can be obtained from a single GC-O run. In the first GC-O run, all volatiles
are detected whose concentrations in the GC effluent are higher than their odor
thresholds. The corresponding volatiles are then characterized by their aroma
quality and intensity as well as by their chromatographic properties, i.e., retention
index (RI). The RI increments, obtained on stationary phases with different polar-
ities, provide additional information about the nature of the aroma-active compo-
nent, such as functional groups. Aroma qualities and intensities are very useful
data for flavorists, who can then use these to create characteristic and complex
aroma notes.
As mentioned in Section II, chromatograms obtained by FID detection and
olfactory response are different. Aroma-active compounds usually do not corre-
spond to the major volatile components in the food. As shown in Fig. 9, many
important odorants of white bread crust were not visible in the gas chromatogram,
for example, 2-acetyl-1-pyrroline (no. 11) (20). This can be explained by the low
odor threshold of these compounds. Identification of such minor components
(Fig. 10) is a challenging task.
Once the aroma-active regions have been selected by a dilution analysis,
the often time-consuming identification experiments can be focused on the most
potent odorants. If further fractionation and clean-up steps are required, GC-O
may again serve as a screening method and guide sample-purification work. This
approach, called sensory-directed chemical analysis, is particularly useful when
identifying unknown compounds with very low threshold values. An impressive
example is the identification of 1-p-menthene-8-thiol as the aroma principle of
grapefruit juice (21). Its threshold is the lowest ever reported for a naturally
occurring compound: 2 10 8 mg/liter water. More examples are listed in Table
2. Most of these compounds occur at very low concentrations and are difficult
to identify using conventional analytical techniques.
In general, the aroma quality of a volatile component combined with more
than one RI value is considered as equivalent to identification by GC-MS. How-
ever, the reference compound should be available because of shifts of RI values,
especially on polar capillaries. The presence of an odorant can be verified by
coelution with the reference compound on capillaries of differing polarity. This
approach is helpful for the identification of odorants with very low threshold
values and unique aroma qualities (Table 3). RI values may also provide crucial
data for the identification of unknown odorants, even if the reference compound
is not available. In some cases, verification by GC-MS is possible by tuning the
detection technique, e.g., looking at typical fragments of the target molecule in
a well-defined region of the gas chromatogram, recording in the SIM mode, and
306 Blank
FIGURE 9 (A) Gas chromatogram and (B) FD chromatogram of the headspace volatiles
of fresh white bread crust. (From Ref. 20.)
FIGURE 10 Chemical structures of the odor-active components identified in the head-

space of fresh white bread crust: 2-methylpropanal (no. 1), diacetyl (no. 2), 3-methylbuta-
nal (no. 4), 2-acetyl-1-pyrroline (no. 11), 1-octene-3-one (no. 12), 2-ethyl-3,5-dimethyl-
pyrazine (no. 13), and (E)-2-nonenal (no. 15). The numbering corresponds to that in
Figure 9.
applying GC-MS/MS. It should be mentioned that, with certain experience, time-

consuming identification work can be limited to a few, still unknown, compounds.
B. Formation of Potent Odorants

The identity of a key aroma compound is a flavor chemists first information.
However, it will not automatically lead to product quality improvement. Addi-
TABLE 2 Aroma Impact Components Newly Identified in Foods

on the Basis of GC-O
Food Compound Ref.
Bread crust, wheat 2-Acetyl-1-pyrroline 22
Beef meat, boiled 2-Methyl-3-furanthiol a 23
Beef meat, roasted 2-Acetyl-2-thiazoline 24
Beef meat, stewed 12-Methyltridecanal a 25
Coffee, roasted 3-Mercapto-3-methylbutyl formatea 26
Cheese (Emmentaler) Furaneol, homofuraneol 27
Grapefruit 1-p-Menthene-8-thiol a 21
Wine (Sauvignon) 4-Mercapto-4-methyl-2-pentanone 28
Tea, green 3-Methyl-2,4-nonanedione 18
Lovage Sotolon 29
a
Odorant was reported for the first time as food constituent.
308 Blank
TABLE 3 Selected Examples for Compounds That Can Be Identified by GC-O on the
Basis of Aroma Quality and Retention Indices
Linear retention indices
Aroma quality
Compound (at sniffing port) SE-54 OV-1701 FFAP
1-Octene-3-one Mushroomlike 982 1065 1315
(Z)-1,5-Octadiene-3-onea Metallic, geraniumlike 985 1085 1395
trans-4,5-Epoxy-(E)-2- Fatty, metallic 1385 1550 1990
decenal a
3-Methyl-2,4-nonanedione a Strawy, haylike 1316 1400 1700
-Decalactone Coconutlike 1685 1733 2185
(E)--Damascenone Honeylike, cooked apples 1395 1500 1825
Sotolon Seasoninglike 1107 1350 2215
Furaneol Caramel-like 1065 1240 2045
2-Isopropyl-3- Earthy, potatolike 1097 1146 1430
methoxypyrazine
2-Isobutyl-3- Earthy, paprikalike 1186 1237 1520
methoxypyrazine
2-Ethyl-3,5- Earthy, roasty 1083 1150 1455
dimethylpyrazine
2,3-Diethyl-5- Earthy, roasty 1155 1219 1485
methylpyrazine
2-Acetyl-1-pyrroline a Roasty 923 1013 1345
2-Acetyl-2-thiazoline Roasty, popcornlike 1110 1245 1720
Methional Cooked potatolike 909 1042 1465
3-Mercapto-2-pentanone Catty, sulfury 907 1018 1365
2-Furfurylthiol Roasty, sulfury 913 1000 1445
2-Methyl-3-furanthiol Meaty, roasty, sweet 870 932 1325
3-Methyl-2-butene-1-thiol a Sulfury, Allium-like, foxy 821 874 1105
3-Mercapto-3-methylbutyl Catty, sulfury 1023 1138 1515
formate a
a
Compound is not commercially available.
tional work on precursors and formation mechanisms is required (Table 4). This
may result in conditions favoring the generation of positive aromas by processing.
It may also support selection of raw materials and give some indications for a
more efficient enzymatic and/or thermal treatment to liberate precursors of key
aroma components.
Recently, furaneol and homofuraneol were detected by GC-O in Maillard
model reactions based on pentoses and different amino acids. This initiated a
systematic study to explain these surprising findings (30). As shown in Fig. 11,
the Strecker aldehydes of glycine and alanine were actively involved in the forma-
TABLE 4 Precursors of Some Aroma Impact Compounds Found in Food
Compounds Food Precursor systems Ref.a

12-Methyltridecanal Stewed beef Plasmaologens 25
2-Methyl-3-furanthiol Boiled beef Thiamine/cysteine (H 2S) 31
2-Ethyl-3,5-dimethylpyrazine Roasted beef, coffee Alanine/methylglyoxal 32
2-Acetyl-2-thiazoline Meatlike model system Cysteamine/methylglyoxal b 33
3-Methyl-2-butene-1-thiol Roasted coffee Prenyl alcohol/H 2S 34
2-Acetyl-1-pyrroline Wheat bread crust Ornithine/methylglyoxal c 35
Furaneol Wheat bread crust Fructose-1,6-bisphosphate 36
Homofuraneol Soy sauce Sedoheptulose-7-phosphate 37
Sotolon Fenugreek 4-Hydroxy-l-isoleucine 38
3-Methyl-2,4-nonanedione Soybeans, green tea Furanoid fatty acids 39
a
Recently published references are preferably cited.
b
2-(1-Hydroxyethyl)-4,5-dihydrothiazole is the key intermediate.
c
1-Pyrroline is an important intermediate.
FIGURE 11 Schematic formation of furaneol and homofuraneol from pentoses (e.g., xy-
lose) in the presence of glycine and alanine elucidated by labeling experiments. The
marked positions () represent 13 C-atoms. (Adapted from Ref. 30.)
310 Blank
tion of furaneol and homofuraneol, respectively. The results were obtained using
the 13 C-labeled precursors and GC-MS/MS as a selective and sensitive detection
technique.
C. Off-Flavor Analysis
GC-O is the method of choice for selecting those components that are responsible
for aroma deviation in food, i.e., an off-flavor. In general, it can be applied to both
foodborne off-flavor formation and off-flavor problems related to contamination.
The latter is caused by odorants that normally do not belong to the overall aroma
of the product, i.e., external contaminants (e.g., packaging) or compounds formed
upon processing and storage (e.g., microbial spoilage). In both cases, the compari-
son of the off-flavor of the contaminated food with the reference product usually
results in a limited number of sensory-relevant compounds, which reflect the differ-
ence in aroma profiles. Identification work can then be focused on these odorants.
Recent work by Spadone et al. (40) on the Rio defect in green coffee from
Brazil and by Marsili et al. (41) on the off-flavor of sugar beet impressively
illustrates the potential of this approach: 2,4,6-trichloroanisole and geosmin were
identified as off-flavor compounds, respectively. Both odorants have very low
odor thresholds in water: 5 10 8 and 5 10 7 mg/liter, respectively. Identifi-
cation work was completed by quantitative data, and the off-flavor activity was
confirmed by sensory evaluation. The approach, based on sensory techniques and
a strong analytical support, provides a good basis for solving off-flavor problems
in a reasonable time (Table 5).
Foodborne off-flavor is mainly caused by concentration shifts in aroma-
active food constituents. This is much more difficult to handle due to the subtle
changes that finally result in an unbalanced aroma. The warmed-over flavor
TABLE 5 Examples of GC-O Recently Applied to Off-Flavors Caused by Direct or

Storage-Related Contamination
Product Off-flavor compound Ref.a
Coffee, green 2,4,6-Trichloroanisole 40

Beet sugar Geosmin, butyric, and isovaleric acids 41
Water, river 2-Methylisoborneol, geosmin 42
Water, mineral C 3-Alkyl benzenes 43
Cognac, cork taint 2,4,6-Trichloroanisole, 2,3,4,6-tetrachlophenol 44
Wine 2-Acetyletetrhydropyridines, furaneol 45,46
Pearl millet, ground, wetted 2-Acetyl-1-pyrroline 47
a
Recently published references are given preference.
TABLE 6 Sensory Evaluation to Check the Representativeness of the Aroma Extract

Obtained from a Commercially Available Food Flavoring with Savory Character
Sample Isolation method Sensory attributes
A (original product) Meaty, roasty, savory, onionlike

B (extract of A) SDE (boiling conditions) Cooked vegetables, burnt, meaty
C (extract of A) SDE (static vacuum at 35C) Meaty, roasty, savory, onionlike
D (extract of SDE residue) a Direct solvent extraction Caramel-like, savory, acidic
a
Sample D was obtained from the SDE residue of sample C by extraction with diethyl ether.
SDE simultaneous distillation-extraction.
Source: Modified from Ref. 52.
(WOF ) of cooked meat is a well-known example in the food industry. Quantita-

tive results of odorants contributing to the off-flavor are indispensable for ob-
taining reliable data about changes in the aroma profile. Using this approach,
hexanal and trans-4,5-epoxy-(E)-2-decenal have recently been found to be the
main contributors of WOF (48).
IV. ANALYTICAL CONSIDERATIONS RELATED TO GC-O

A. Representativeness of the Aroma Extract
An excellent review of sample preparation has recently been published by Tera-
nishi and Kint (49). In general, heat treatment should be limited to avoid forma-
tion of artefacts and decomposition of aroma impact components. Enzymatic ac-
tivity in natural products is another critical parameter that should be controlled
during sample preparation. In general, there is no ideal extraction method in food
aroma analysis. The choice of an extraction procedure depends on the food and
is always a compromise.
The issue of representativeness of the aroma extract has recently been dis-
cussed in detail by Etievant and coworkers (50,51). Apparently, little attention is
paid in the literature to the quality of the aroma extract, although it is well known
that aroma composition depends on the extraction method used. Indeed, the first
objective in aroma analysis is to ensure that the extract is representative of the
original product. The conditions of extraction and concentration should be designed
in such a way as to obtain a sample with an authentic aroma. To make sure that
an aroma extract merits further analytical and sensory characterization, it is highly
recommended to check its representativeness, e.g., using triangle, similarity scaling,
descriptive, or matching tests. This is the basis for obtaining reliable results.
As illustrated in Table 6, the isolation method used to prepare sample B
changed the unique meaty/savory note of the original product (sample A) to
312 Blank
boiled vegetables, grilled, burnt. In contrast, the isolation method used to ob-
tain sample C resulted in an extract that revealed the authentic aroma of the
original product. Consequently, this extract was further characterized by GC-O
and other instrumental and analytical techniques. The results and the isolation
techniques used will be discussed in the following section.
B. Comparison of Different Isolation Techniques

1. SDE Under Reduced Pressure
The simultaneous distillation-extraction (SDE) under atmospheric pressure (53)
is not always the most appropriate technique, and its use should be carefully
considered. This technique is an elegant and rapid extraction method resulting in
an aroma extract that is ready to be injected into a GC system after concentration.
However, heat-induced artefact formation, decomposition of labile compounds,
and loss of very volatile compounds are serious drawbacks. Furthermore, only
steam-distillable volatiles are extracted. Polar compounds, such as hydroxyfura-
nones and phenols, are particularly poorly recovered.
Considerable effort has been made to overcome at least one of the limitations
of the SDE technique, i.e., heat-induced changes of the aroma extract. A modified
SDE apparatus was designed to work under static vacuum (SDE-SV), thus allowing
extraction at 3035C (54). Several solvents were tested, of which butylethyl ether
showed good results for various classes of substances (52). The extract (SDE frac-
tion) can be directly analyzed by GC without any concentration step. SDE-SV is
more time-consuming than conventional SDE and also more delicate in handling:
exact control of three temperatures (aqueous sample, organic solvent, cooling by
cryostat) is necessary, but results in a ready-to-inject aroma extract free of artefacts.
GC-O of the SDE-SV extract of a commercial meaty/savory flavoring
(sample C, Table 6) revealed 15 odor-active components out of about 100 vola-
tiles. However, only six odorants showed FD factors of 2 8 and higher (Table 7).
The meaty/savory note was mainly imparted by odorants containing sulfur. The
cis-isomer of 2-methyl-3-tetrahydrofuranthiol was also found, but did not contrib-
ute to the overall aroma. Identification was mainly based on GC-MS and NMR
and was verified by commercially available or synthesized reference compounds.
These are essential for unequivocal identification. The chemical structures of the
aroma impact compounds identified in the flavoring are shown in Fig. 12.
2. SDE Combined with Direct Solvent Extraction

As shown in Table 8, the acidic fraction (sample D in Table 6) obtained from the
residue after extraction by SDE-SV was a good source of additional information,
particularly about polar aroma compounds. They were isolated from the SDE
residue by direct solvent extraction with diethyl ether, purified by extraction with
TABLE 7 Odorants Identified on the Basis of AEDA in the SDE-SV Extract (Sample
C) of a Commercially Available Food Flavoring with Savory Character
Retention index
Odor quality FD factor
No. Compound a OV-1701 FFAP (GC-O) (2 n)
1 2-Methyl-3-furanthiol 931 1325 Meaty, roasty, sweet 1011
(MFT)
2 trans-2-Methyl-3- 992 1315 Meaty, savory, onion 1011
tetrahydrofuranthiol
3 2-Furfurylthiol 1000 1450 Sulfury, roasty 1415
4 Methional 1044 1465 Cooked potato, boiled 1011
5 S-(2-Methyl-3-furyl)- 1238 1700 Meaty, roasty 1011
ethanethioate
6 4-Acetyloxy-2,5- 1430 2005 Caramel-like, savory 89
dimethyl-3(2H)-
furanone
a
Compounds nos. 16 were detected in the SDE fraction obtained under static vacuum.
FIGURE 12 Aroma impact compounds identified in a meaty/savory food flavoring: 3-

methyl-2-furanthiol (no. 1), trans-2-methyl-3-tetrahydrofuranthiol (no. 2), 2-furfurylthiol
(no. 3), methional (no. 4), S-(2-methyl-3-furyl)-ethanethioate (no. 5), 4-acetyloxy-2,5-
dimethyl-3(2H)-furanone (no. 6), furaneol (no. 8), and sotolon (no. 9). Acetic acid (no. 7)
is not shown. The numbering corresponds to that in Tables 7 through 9.
314 Blank
TABLE 8 Odorants Identified on the Basis of AEDA in the Acidic Fraction (Sample
D) of a Commercially Available Food Flavoring with Savory Character
Retention index
Odor quality FD factor
No. Compound a OV-1701 FFAP (GC-O) (2 n)
7 Acetic acid 785 1460 Acetic, pungent 1011
8 Furaneol 1240 2045 Caramel-like, sweet 1617
9 Sotolon 1350 2220 Seasoninglike 1112
a
Compounds nos. 79 were detected in the acidic fraction (SDE residue of sample C).
sodium carbonate (0.5 mol/liter) and after acidification reextracted with the sol-
vent. Furaneol (no. 8) was the dominating odorant in this extract; accordingly,
sample D was mainly described as caramel-like.
3. SDE Versus Static Headspace

As mentioned earlier (Sec. II.C), headspace GC analysis yields additional data
about very volatile compounds, which are usually lost during conventional sam-
ple preparation. The sensory relevance of odorants present in the headspace above
a food can be evaluated by combining AEDA with the static headspace technique
(18). This new approach, called static headspace GC-O, was applied to the food
flavoring discussed above.
The odorants listed in Table 9 were identified on the basis of their chro-
matographic and sensory properties using the same analytical conditions as ap-
plied to AEDA. A stepwise reduction of the headspace volume revealed the most
potent odorants. The medium, dry or aqueous, significantly influenced the results.
For example, thiols nos. 1 and 2 were completely lacking in the headspace of
the solid product, whereas in the aqueous medium they showed high odor poten-
cies. These compounds could have been either efficiently encapsulated or liber-
ated from nonvolatile precursors by hydrolysis.
Compounds 14 and 1013 were more abundant in the headspace of the
aqueous sample. Others dominated in the headspace of the dry product, such as
the polar and well watersoluble compounds 68. Diacetyl and the lipid degrada-
tion products 1113 contributed more significantly to the aroma of the headspace
than the liquid aroma extract obtained by SDE-SV.
4. Vacuum Distillation Versus Direct Solvent Extraction

The analysis of furaneol is an excellent example to illustrate that, unless appro-
priate isolation techniques are used, an important odorant may be missed. Contra-
TABLE 9 Static Headspace GC-O of a Commercially Available Food Flavoring with

Savory Aroma Character
FD factor (2 n) b
Aroma quality Solid Aqueous
No. Compound a (on GC-O) sample sample
12 MFT trans-2-methyl-3-tetra- Meaty, savory 8
hydrofuranthiol
3 2-Furfurylthiol Sulfury, roasty 4 8
4 Methional Cooked potato 4 8
5 S-(2-Methyl-3-furyl)-ethanethioate Meaty, roasty 2 2
6 4-Acetyloxy-2,5-dimethyl-3(2H)- Caramel-like 5 2
furanone
7 Acetic acid Acidic, pungent 2
8 Furaneol Caramel-like 2
9 Sotolon Seasoninglike
10 Diacetyl Buttery, sweet 2 4
11 Hexanal Green, fatty 2 4
12 1-Octene-3-one Mushroomlike 3 7
13 (E)-2-Nonenal Fatty 2 7
a
Numbering corresponds to that in Tables 7 and 8.
b
The initial headspace volume of 20 ml was set as FD 1. The headspace volume was stepwise
reduced and the GC-O procedure repeated to 0.1 ml, which corresponds to FD 2 8.
dictory results have been published concerning the occurrence and concentration
of furaneol and its methylether (MDMF) in strawberries. Therefore, no clear
conclusion can be drawn about their sensory relevance. As shown in Table 10,
it was rather difficult to detect furaneol in vacuum distillates (Refs. A and B). It
is highly oxygenated and, therefore, does not steam-distill due to its low vapor
pressure in aqueous samples (58). Consequently, furaneol must be extracted with
solvent (Refs. C and D). Cold on-column injection should ideally be used to
avoid thermally induced decomposition of furaneol (59) (see Sec. IV.C).
C. Optimized Chromatographic Conditions

1. Aroma Alteration Prior to Chromatography
As many odorants are rather labile and occur at low concentrations, a long storage
period between sample preparation and GC-O should be avoided. The choice of
solvent is another critical parameter. Certain thiols rapidly dimerize in diethyl
ether upon refrigerated storage (60), for example, 2-methyl-3-furanthiol, which
316 Blank
TABLE 10 Results Reported in the Literature on the Presence of Furaneol and Its
Methylether in StrawberriesEffect of Analytical Conditions
Analytical parameter Ref. A (56) Ref. B (57) Ref. C (58) Ref. D (59)
Extraction
Vacuum distillation x x
Direct solvent extraction x x
Injection mode
Split/Splitless x x x
Cold on-column x
Concentration (mg/kg)
Furaneol 0.01 2.26.3 2.716.2
MDMF 0.2 0.12.6 0.510.9 Not determined
is an aroma impact compound of boiled beef (23). In such cases, the aroma con-
centrate should preferably be stored in pentane at 30C, if possible under an
inert gas, to avoid alteration of the aroma profile.
Furthermore, the aroma extract should be injected using the cold on-column
technique. Unstable volatiles readily decompose in a heated injector block and
form artefacts, e.g., hydroxyfuranones and thiols. Sulfur-containing compounds
are particularly susceptible to heat-induced decomposition that can take place
during split/splitless injection, GC separation, or in the GC-MS interface (61).
Indeed, many newly reported constituents in Allium chemistry are artefacts (Fig.
13). HPLC and low-temperature GC and GC-MS conditions have been proposed
for their analysis (62,63).
2. Effect of Chromatography on GC-O Data

The problems involved in analyzing very low amounts of often labile components
should not be neglected in GC-O (64). Testing column quality on a regular basis
is indispensable. Several mixtures are commercially available for testing polar
and apolar capillaries. The so-called Grob test is highly recommended, as it rap-
idly indicates the quality of the analytical capillary (65). For example, adsorption
effects due to active surfaces are indicated by tailing of the 1-octanol peak. The
quality of an FFAP column can be tested by injecting mixtures of alkanes and
free fatty acids.
Many products, especially heat-processed and fermented foods, contain a
large variety of substance classes. Therefore, an aroma extract should be analyzed
on at least two capillaries of different polarity: an apolar phase (e.g., OV-1, SE-
54) and a polar phase (e.g., Carbowax, FFAP). This may help to achieve a better
separation of odor-active compounds, as shown in Fig. 3. The medium polar capil-
FIGURE 13 Formation of artefacts in Allium chemistry. (A) Allicin readily decomposes

in a heated injector block forming two thioacrolein isomers before GC separation (62).
(B) Bis-(1-propenyl)-disulfide rearranges at 85C to thienyl compounds commonly found
in Allium distillates (63).
lary OV-1701 is a good compromise for analyzing both apolar and rather polar
compounds. In general, chromatography may affect the FD factor and Charm value,
particularly at high dilution levels when picogram amounts are analyzed.
As shown in Table 11, odor thresholds determined by GC-O may vary by
several orders of magnitude depending on the stationary phase used. Consequently,
such effects will also influence the FD factor and Charm value because they repre-
TABLE 11 Odor Thresholds a (ng/liter air) of Some Selected Odorants as Affected by

the Stationary Phase b
Compound SE-54 OV-1701 FFAP
2-Methyl-3-furanthiol (MFT) 0.0010.002 n.d. 510

Abhexon 24 n.d. 0.0020.004
Sotolon n.d. 0.61.2 0.010.02
Furaneol n.d. 12 0.51.5
3,4-Dimethylcyclopentenolone n.d. 12 0.050.1
Cyclotene n.d. 1020 1020
a
Odor thresholds were determined by GC-O (14) using (E)-2-decenal as internal standard.
b
Capillaries were selected using the Grob test.
318 Blank
FIGURE 14 Gas chromatography of 4-methoxy-2,5-dimethyl-3(2H)-furanone (no. 1) and

furaneol (2) on SE-54 and FFAP fused silica capillaries (I. Blank and W. Grosch, unpub-
lished results). The original sample (FD-1) was stepwise diluted (FD-2, FD-4, FD-8) and
analyzed using the same conditions (injection: cold on-column).
sent the odor threshold of the compound at a given concentration. Indeed, different
FD factors were determined for MFT on SE-54 and FFAP: 2 14 and 2 6, respectively.
On the contrary, abhexon showed higher FD factors on FFAP than SE-54: 2 16
and 2 5, respectively. Consequently, FD factors should be determined on suitable
capillaries (64). Compounds with low threshold values are much more affected by
this phenomenon, i.e., sotolon compared to furaneol. This can be explained by a
chromatographic discrimination at low concentration as discussed below.
3. Discrimination of Odorants on Stationary Phases

The phenomenon of adsorption/instability of enoloxo compounds during GC
analysis is illustrated in Fig. 14. Furaneol and its methylether were analyzed at
different concentrations by diluting the samples 1 : 1, i.e., FD-1, FD-2, FD-4, and
FD-8. The chromatography of 4-methoxy-2,5-dimethyl-3(2H)-furanone (no. 1)
was not affected by the stationary phase. On the contrary, furaneol (no. 2) was
partially lost on SE-54. Chromatographic behavior on FFAP was acceptable.
FIGURE 15 Yields of cyclotene, 2-methoxy-3-methyl-2-cyclopentene-1-one, furaneol,

and 4-methoxy-2,5-dimethyl-3(2H)-furanone as a function of different polar stationary
phases on GC with fused silica capillaries (injection: cold on-column). (Adapted from
Ref. 64.)
320 Blank
Discrimination of several compounds at low concentrations on different

stationary phases was studied using furaneol, cyclotene, and their methylethers
(Fig. 15). Decreasing amounts were injected via cold on-column, and the yields
were determined by projecting the peak areas onto that of the undiluted sample,
which was set at 100%. These experiments were continued until the detection
limit of the FID was reached. FFAP was found to be the most suitable stationary
phase for the analysis of polar compounds, followed by Carbowax. On the con-
trary, yields on SE-54 and OV-1701 decreased strongly with increasing dilution.
The corresponding methylethers showed much better chromatographic proper-
ties, most likely due to the blocked hydroxyl group, which reduces interaction
with the stationary phase.
About 1 ng of furaneol injected on-column onto an FFAP still revealed a
symmetrical peak by FID detection that can be integrated for quantification. It
has been reported that, using a heated injector block, 10 ng of furaneol still gives
a well-defined, sharp peak on a Carbowax fused silica capillary (66). However,
we found that about 2030 ng of furaneol on an OV-1701 resulted in an almost
undetectable broad peak. Therefore, the FD factors of furaneol determined on
FFAP are usually higher than those on OV-1701. The same phenomenon was
observed for homofuraneol, sotolon, and abhexon, which are potent odorants and
contribute to the flavor of several heat-processed and fermented foods.
V. SENSORY ASPECT OF GC-O

A. Role of Odor Thresholds in GC-O
It is very useful to have an approximate idea about the threshold value of odor-
ants. The odor threshold of a compound (O x), measured as ng/liter of air, can
be determined by GC-O using (E)-2-decenal as sensory internal standard ac-
cording to the following equation (14):
OI DI Cx
Ox (1)
CI Dx
where C I and C x represent the concentrations and DI and D x the dilution values
of the internal standard and the odorant, respectively. The term O I is the odor
threshold of the internal standard, (E)-2-decenal, which has previously been de-
termined: 2.7 ng/liter air (67). This compound must be present in the solution
containing the odorant(s). Therefore, all thresholds listed in Table 12 are related
to the sensory internal standard, which allows an objective comparison of the
values.
The information about odor thresholds determined by GC-O can be of great
help in identifying sensory-relevant compounds of both positive and off-flavors,
TABLE 12 Odor Thresholds Determined by GC-O a of

Some Aroma Impact Components b Found in
Thermally Processed Foods c
Odor threshold
Compound (ng/liter air)
trans-4,5-Epoxy-(E)-2-decenal 0.00050.005
(Z)-2-Nonenal 0.0020.008
(E)-2-Nonenal 0.040.16
(E,E)-2,4-Decadienal 0.050.2
(E,Z)-2,6-Nonadienal 0.10.4
Hexanal 1545
(E)--Damascenone 0.0020.004
(Z)-1,5-Octadiene-3-one 0.0030.006
3-Methyl-2,4-nonanedione d 0.0070.014
1-Octene-3-one 0.050.1
4-Methylacetophenone 24
Diacetyl 1020
4-Ethylguaiacol 0.010.03
4-Vinylguaiacol 0.40.8
Eugenol 0.20.4
Vanillin 0.61.2
4-Methylphenol 0.31
Myristicin 12
2-Isopropyl-3-methoxypyrazine 0.00050.001
2,3-Diethyl-5-methylpyrazine 0.0090.018
2-Ethyl-3,5-dimethylpyrazine 0.0070.014
2-Isobutyl-3-methoxypyrazine 0.0020.004
2-Acetyl-1-pyrroline 0.020.04
2-Acetyltetrahydropyridine 0.10.2
3-Mercapto-3-methylbutyl formate 0.00020.0004
2-Methyl-3-furanthiol e 0.0010.002
2-Furfurylthiol 0.010.02
Dimethyltrisulfide 0.060.12
Methional 0.10.2
2-Acetylthiazol 25
a
Odor thresholds were determined by GC-O (14) on an OV-
1701 using (E)-2-decenal as sensory internal standard.
b
From Refs. 32, 64, 6870.
c
The threshold values of enoloxo compounds determined on
FFAP are listed in Table 11, i.e., abhexon, sotolon, furaneol,
cyclotene, and 3,4-dimethylcyclo-pentenolone.
d
Odor threshold was determined on a Carbowax.
e
Odor threshold was determined on an SE-54.
322 Blank
particularly in case of separation problems (peak overlapping) or similar mass spec-

tra. 2-Ethyl-3,5-dimethylpyrazine and 2-ethyl-3,6-dimethyl-pyrazine have similar
mass spectra but can easily be distinguished based on their odor thresholds: 0.007
0.014 and 2.55 ng/liter air, respectively (32). Although 1-octene-3-ol and 1-oc-
tene-3-one coelute on apolar capillaries, the presence of the latter can be verified by
dilution of the sample due to its 100-fold lower threshold. Thus, dilution techniques
provide additional data for positive identification. Compounds with odor thresholds
lower than 1 ng/liter air are usually below the detection limit of an FID.
B. Limitations of the GC-O Approach

The importance of the representativeness of the aroma extract and the possible
effects of the GC analysis on GC-O data have already been discussed in detail
(see Sec. IV). Several recently published articles are recommended regarding
practical aspects of the GC-O procedure (12,13,7173). Therefore, only a few
additional remarks will be made here.
All GC-O runs of a dilution analysis should be performed within one week
to reduce variability in GC analysis and sensorial perception. If analysis takes
longer to complete, assessors can have gaps during sniffing, i.e., they do not
detect a substance at a certain dilution, but detect it again at a higher dilution (73).
The first step is to establish a profile of the original aroma extract by detecting the
odor-active regions and describing their aroma characteristics. This should be
done on two capillaries of different polarity. A medium polar capillary is pro-
posed for the dilution analysis. All odor-active regions detected in the original
aroma extract should be sniffed throughout the entire dilution series, consisting
of not more than 10 samples. The last 5 dilutions should be repeated on the
second capillary, preferably an FFAP. Some of the dilutions should be evaluated
by additional assessors. The use of humidified air is recommended to reduce
olfactory fatigue by nasal dehydration. Consider that the number of odorants
detectable by GC-O depends on the extraction method and the threshold of the
volatiles, but also on parameters that are arbitrarily selected, i.e., amount of food
sample, concentration factor, sample volume injected.
Problems such as representativeness, GC analysis, and gaps during
sniffing can be solved by taking the necessary precautions. More serious limita-
tions of the GC-O approach originate from the sensory area. Methods based on
odor threshold detection, i.e., GC-O using dilution techniques (Charm, AEDA)
and the OAV concept (ratio of concentration to threshold), are not consistent
with psychophysical views (74). The major problems are that thresholds vary
depending on the experimental conditions, differing intensity functions for vola-
tiles above the threshold are not accommodated, and no prediction about the
activity of volatiles in a mixture is possible, especially if they occur at concentra-
tions below threshold.
Differing intensity functions for volatiles account for a well-known phe-

nomenon in GC-O: some intensely smelling compounds disappear after a few
dilution steps (e.g., vanillin), whereas others with a lower aroma intensity in the
original extract have the highest FD factors. The sensory relevance of the latter
is overestimated. (E)--Damascenone is a typical representative for such com-
pounds, which are characterized by a relatively flat dose/intensity function. This
is most likely why (E)--damascenone does not play a major role in the aroma
of coffee, despite low threshold values (Table 12), i.e., high FD factors and
OAVs. In other words, threshold concentration does not necessarily correlate
with aroma potency.
A more satisfactory but more difficult approach is to provide intensity mea-
sures, which can only be carried out with a well-trained panel (75). A fundamental
weakness of all of these techniques is that they do not account for interactions
arising in the olfactory system or between taste and smell. The chemical bases
of these senses are still not sufficiently understood.
C. Interpretation of GC-O Data

Properly performed GC-O and adequate knowledge about the possible limitations
of the technique are the basis for a realistic interpretation of the results. The
flavor scientists major questions are What can be concluded from GC-O data?
and Where is the limit of a realistic interpretation?
Certainly, GC-O is a first essential step for distinguishing odor-active com-
pounds from volatiles without odor impact. This screening procedure is the basis
for identification experiments. GC-O also provides a first indication about the
odor potency of volatile compounds, i.e., to what extent they individually contrib-
ute to the overall aroma. However, in most cases, final conclusions about their
sensory relevance cannot be drawn and further work is necessary (see Sec. VI).
Analytical and sensorial data cannot be presented with the same precision.
While RI values and mass spectra can be precisely determined, GC-O data lack
comparable accuracy and reproducibility. FD factors and Charm values are ap-
proximations of the sensory relevance of an odorant. In fact, a 256-fold dilution
is a rough estimation, depending on extraction yields and the assessor, and could
also be 128 or 512. Such exact values are rather misleading: GC-O techniques
are not this accurate. The use of 2 n , where n is the number of dilution steps, may
help to avoid overinterpretation of GC-O data (76). Though 256 and 2 8 represent
the same value, the latter gives a more realistic idea of the odor potency of a
compound. It should be mentioned, however, that FD factors do not normally
differ by more than two dilution steps.
Furthermore, Charm or FD chromatograms can be divided into three re-
gions, represented by compounds with high, medium, or low dilution values. This
classification of odorants could be used as selection criteria for identification,
324 Blank
which would be focused on the first two categories, i.e., odorants of high and
medium potencies. Less effort would be attributed to those odorants contributing
to the background aroma. However, the role of the aroma quality should not
be neglected in this context: several background odorants with a typical note may
also contribute to the overall aroma.
The approach presented above could be standardized by setting the highest
dilution value at 2 10 ( 1024) and relating the remaining values accordingly.
Odorants of high and medium potencies would be grouped depending on their
dilution values, i.e., 2 810 and 2 57, respectively. In this way, the role of an odorant
in different foods could easily be estimated. Moreover, it would allow a better
comparison of GC-O data from different laboratories.
In summary, GC-O techniques should be seen as screening methods to gain
an insight into important contributors to a characteristic aroma (7,71). GC-O
performed as Charm analysis and Osme have also been claimed as quantitative
bioassays (10,77). However, more time is needed for training of assessors and
verification using statistical means.
VI. OUTLOOK
The aim of GC-O techniques in food aroma research is to determine the relative
odor potency of compounds present in the aroma extract. This method gives the
order of priority for identification and thus indicates the chemical origin of olfac-
tory differences (7). The value of the results obtained by GC-O depends directly
on the effort invested in sample preparation and analytical conditions. Analysis
of an aroma extract by dilution techniques (AEDA, Charm) combined with static
headspace GC-O provides a complete characterization of the qualitative aroma
composition of a food. However, this is only the first step in understanding the
complex aroma of a food.
State of the art in food aroma research today is based on a combined sensory
and analytical approach. It is basically composed of the following three steps, which
can be applied to the characterization of both positive aroma and off-flavors:
1. Qualitative aroma composition (based on GC-Olfactometry)
2. Quantitative aroma composition (odor activity value concept)
3. Aroma recombination studies (aroma simulation based on quantitative
data)
Work starts with the analysis of the aroma composition and is completed when
the aroma of the food can be simulated in an appropriate matrix on the basis of
the quantitative data obtained. The last step is essential and validates the analyti-
cal results (78). Recently published data on stewed beef (79) and coffee brew
(80) impressively demonstrate the potential of this approach. However, a crucial
step is accurate quantification of the aroma impact compounds. Special tech-

niques are necessary to quantify labile odorants at low concentrations, i.e., isotope
dilution assay using the labeled odorant as internal standard. A major concern is
the availability of these labeled compounds. The recently published review arti-
cles of Grosch (81) and Schieberle (72) are recommended for more details.
In conclusion, there is a clear need to improve the quality and stability of
food aromas and flavors. The techniques presented above represent an attractive
approach for analyzing aromas more purposefully. Depending on the case, it is
possible to simplify the approach and find compromises in all three phases so
that essential work can be done in a reasonable time. The food and flavor industry
is well advised to profit from this development and to adapt the different tech-
niques to their specific needs.
ACKNOWLEDGMENTS
The author is grateful to Drs. E. Prior, D. Roberts, L. Fay, and J. Loliger for
their helpful suggestions in preparing this manuscript.
APPENDIX: ABBREVIATIONS AND TRADE NAMES

Chemicals
Abhexon 5-Ethyl-3-hydroxy-4-methyl-2(5H)-furanone
Cyclotene 2-Hydroxy-3-methyl-2-cyclopentene-1-one
(E)--Damascenone 1-(2,6,6-Trimethyl-1,3-cyclohexadienyl)-(E)-2-
butene-1-one
3,4-Dimethylcyclopentenolone 2-Hydroxy-3,4-dimethyl-2-cyclopentene-1-one
4-Ethylguaiacol 4-Ethyl-2-methoxyphenol
Eugenol 4-(1-Propenyl)-2-methoxyphenol
Furaneol 4-Hydroxy-2,5-dimethyl-3(2H)-furanone
Furaneol is a trade name of Firmenich,
Geneva, Switzerland
Homofuraneol 2(5)-Ethyl-4-hydroxy-5(2)-methyl-3(2H)-
furanone
Isovaleric acid 3-Methylbutanoic acid
MDMF 4-Methoxy-2,5-dimethyl-3(2H)-furanone
Methional 3-Methylthio-1-propanal
Methylglyoxal 2-Oxopropanal
MFT 2-Methyl-3-furanthiol
Sotolon 3-Hydroxy-4,5-dimethyl-2(5H)-furanone
4-Vinylguaiacol 4-Vinyl-2-methoxyphenol
326 Blank
Analytical Techniques
AEDA Aroma Extract Dilution Analysis
Charm Combined Hedonic and Response Measurement
FID Flame Ionization Detector
FFAP Free Fatty Acid Phase (polar stationary phase for GC)
GC Gas Chromatography (using capillary columns)
GC-O GCOlfactometry
GC-MS GCMass Spectrometry
GC-MS/MS GCTandem-MS
HPLC High-Performance Liquid Chromatography
MS Mass Spectrometry
OAV Odor Activity Value (ratio of concentration to odor threshold)
Osme From the Greek word , meaning smell
OV-101 Ohio Valley apolar stationary phase for GC
OV-1701 Ohio Valley medium polar stationary phase for GC
RI Retention Index
SDE Simultaneous Distillation Extraction
SDE-SV SDE under Static Vacuum
SE-54 Apolar stationary phase for GC
SIM Selective Ion Monitoring (GC-MS technique)
REFERENCES
1. L. M. Nijssen, L. M. Visscher, H. Maarse, L. C. Willemsens, and M. H. Boelens

(eds.), Volatile compounds in food. Qualitative and quantitative data, TNO Nutrition
and Food Research Institute, 7th ed., Zeist, 1997.
2. G. H. Fuller, R. Steltenkamp, and G. A. Tisserand, The gas chromatograph with
human sensor: perfumer model, Ann. NY Acad. Sci. 116:711 (1964).
3. M. Rothe and B. Thomas, Aromastoffe des Brotes, Z. Lebensm. Unters. Forsch.
119:302 (1963) (in German).
4. D. G. Guadagni, R. G. Buttery, and J. Harris, Odour intensities of hop oil compo-
nents, J. Sci. Food Agric. 17:142 (1966).
5. M. C. Meilgaard, Aroma volatiles in beer: purification, flavor threshold and interac-
tion, Geruch- und Geschmacksstoffe (F. Drawert, ed.), Verlag Hans Carl, Nurnberg,
1975, p. 210.
6. T. E. Acree, J. Barnard, and D. G. Cunningham, A procedure for the sensory analysis
of gas chromatographic effluents, Food Chem. 14:273 (1984).
7. W. Grosch, Detection of potent odorants in foods by aroma extract dilution analysis,
Trends Food Sci. Technol. 4:68 (1993).
8. H. Maarse and D. G. van der Heij (eds.), Trends in Flavour Research, Elsevier Sci-
ence, Amsterdam, 1994, pp. 179190, 191209, 211220, 267270, 271276,
305309, 345351, 409415.
9. M. R. McDaniel, R. Miranda-Lopez, B. T. Watson, N. J. Micheals, and L. M. Libbey,
Pinot noir aroma: a sensory/gas chromatographic approach, Flavors and Off-Flavors

(G. Charalambous, ed.), Elsevier, Amsterdam, 1990, p. 23.
10. M. A. A. P. da Silva, D. S. Lundahl, and M. R. McDaniel, The capability and psycho-
physics of Osme: a new GC-olfactometry technique, Trends in Flavour Research
(H. Maarse and D. G. van der Heij, eds.), Elsevier, Amsterdam, 1994, p. 191.
11. H. Guichard, E. Guichard, D. Langlois, S. Issanchou, and N. Abbott, GC sniffing
analysis: olfactive intensity measurements by two methods, Z. Lebensm. Unters.
Forsch. 201:344350 (1995).
12. T. E. Acree, Bioassays for flavor, Flavor Science. Sensible Principles and Tech-
niques (T. E. Acree and R. Teranishi, eds.), ACS Professional Reference Book,
American Chemical Society, Washington, DC, 1993, p. 1.
13. T. E. Acree, Gas chromatography-olfactometry, Flavor Measurement (C. T. Ho and
C. H. Manley, eds.), Marcel Dekker, New York, 1993, p. 77.
14. F. Ullrich and W. Grosch, Identification of the most intense odor compounds formed
during autoxidation of linoleic acid, Z. Lebensm. Unters. Forsch. 184:277 (1987).
15. W. Grosch and P. Schieberle, Bread flavour: qualitative and quantitative analysis,
Characterization, production and application of food flavours, Proceedings of 2nd
Wartburg Aroma Symp., Akademie-Verlag, Berlin, 1987, p. 139.
16. D. D. Roberts and T. E. Acree, Gas chromatography-olfactometry of glucose-proline
Maillard reaction products, Thermally Generated Flavors (T. H. Parliment, M. J.
Morello, and R. J. McGorrin, eds.), American Chemical Society, Washington, DC,
1994, p. 71.
17. I. Blank, A. Sen, and W. Grosch, Aroma impact compounds of Arabica and Robusta
coffee. Qualitative and quantitative investigations, Proc. 14th Int. Conf. Coffee Sci.
(ASIC 14), San Francisco, 1992, p. 117.
18. H. Guth and W. Grosch, Identification of potent odourants in static headspace sam-
ples of green and black tea powders on the basis of aroma extract dilution analysis
(AEDA), Flav. Fragr. J. 8:173 (1993).
19. P. Semmelroch and W. Grosch, Analysis of roasted coffee powders and brews by
gas chromatography-olfactometry of headspace samples, Lebensm. Wiss. Technol.
28:310 (1995).
20. P. Schieberle and W. Grosch, Changes in the concentration of potent crust odourants
during storage of white bread, Flav. Fragr. J. 7:213 (1992).
21. E. Demole, P. Enggist, and G. Ohloff, 1-p-Menthene-8-thiol: a powerful flavor im-
pact constituent of grapefruit juice (Citrus paradisi MacFayden), Helv. Chim. Acta
65:1785 (1982).
22. P. Schieberle and W. Grosch, Identification of the volatile flavour compounds of
wheat bread crustcomparison with rye bread crust, Z. Lebensm. Unters. Forsch.
180:474 (1985).
23. U. Gasser and W. Grosch, Identification of volatile flavour compounds with high
aroma values from cooked beef, Z. Lebensm. Unters. Forsch. 186:489 (1988).
24. C. Cerny and W. Grosch, Evaluation of potent odorants in roasted beef by aroma
extract dilution analysis, Z. Lebensm. Unters. Forsch. 194:322 (1992).
25. H. Guth and W. Grosch, 12-Methyltridecanal, a species-specific odorant of stewed
beef, Lebensm. Wiss. Technol. 26:171 (1993).
26. W. Holscher, O. G. Vitzthum, and H. Steinhart, Identification and sensorial evalua-
328 Blank
tion of aroma-impact-compounds in roasted Colombian coffee, Cafe, Cacao, The

34:205 (1990).
27. M. Preininger and W. Grosch, Evaluation of key odorants of the neutral volatiles
of Emmentaler cheese by the calculation of odour activity values, Lebensm. Wiss.
Technol. 27:237 (1994).
28. P. Darriet, T. Tominaga, V. Lavigne, J. N. Boidron, and D. Dubourdieu, Identifica-
tion of a powerful aromatic component of Vitis vinifera L. var. Sauvignon wines:
4-mercapto-4-methyl-2-pentanone, Flav. Fragr. J. 10:385 (1995).
29. I. Blank and P. Schieberle, Analysis of the seasoning-like flavour substances of a
commercial lovage extract (Levisticum officinale Koch), Flav. Fragr. J. 8:191
(1993).
30. I. Blank and L. B. Fay, Formation of 4-hydroxy-2,5-dimethyl-3(2H)-furanone and
4-hydroxy-2(or 5)-ethyl-5(or 2)-methyl-3(2H)-furanone through Maillard reaction
based on pentose sugars, J. Agric. Food Chem. 44:531 (1996).
31. W. Grosch and G. Zeiler-Hilgart, Formation of meat-like flavor compounds, Flavor
PrecursorsThermal and Enzymatic Conversations (R. Teranishi, G. R. Takeoka,
and M. Guntert, eds.), Washington, DC, 1992, p. 183.
32. C. Cerny and W. Grosch, Precursors of ethyldimethylpyrazine isomers and 2,3-
diethyl-5-methylpyrazine formed in roasted beef, Z. Lebensm. Unters. Forsch. 198:
210 (1994).
33. T. Hofmann and P. Schieberle, Studies on the formation and stability of the roast-
flavor compound 2-acetyl-2-thiazoline, J. Agric. Food Chem. 43:2946 (1995).
34. W. Holscher, O. G. Vitzthum, and H. Steinhart, Prenyl alcoholsource of odorants
in roasted coffee, Agric. Food Chem. 40:655 (1992).
35. P. Schieberle, The role of free amino acids present in yeast as precursors of the
odorants 2-acetyl-1-pyrroline and 2-acetyltetrahydropyridine in wheat bread crust,
Z. Lebensm. Unters. Forsch. 191:206 (1990).
36. P. Schieberle, Studies on the formation of furaneol in heat-processed foods, Flavor
PrecursorsThermal and Enzymatic Conversations (R. Teranishi, G. R. Takeoka,
and M. Guntert, eds.), Washington, DC, 1992, p. 164.
37. M. Sasaki, N. Nunomura, and T. Matsudo, Biosynthesis of 4-hydroxy-2(or 5)-ethyl-
5(or 2)-methyl-3(2H)-furanone by yeasts, J. Agric. Food Chem. 39:934 (1991).
38. I. Blank, J. Lin, L. B. Fay, and R. Fumeaux, Formation of 3-hydroxy-4,5-dimethyl-
2(5H)-furanone (sotolon) from 4-hydroxy-L-isoleucine, Bioflavour 95 (P. X. Etie-
vant and P. Schreier, eds.), INRA Editions (Les Colloques, no. 75), Paris, 1995,
p. 385.
39. W. Grosch, U. C. Konopka, and H. Guth, Characterization of off-flavors by aroma
extract dilution analysis, Lipid Oxidation in Food (A. J. St. Angelo, ed.), American
Chemical Society, Washington, DC, 1992, p. 266.
40. J. C. Spadone, G. Takeoka, and R. Liardon, Analytical investigation of Rio off-
flavor in green coffee, J. Agric. Food Chem. 38:226 (1990).
41. R. T. Marsili, N. Miller, G. J. Kilmer, and R. E. Simmons, Identification and quan-
tification of the primary chemicals responsible for the characteristic malodor of
beet sugar by purge and trap GC-MS-OD techniques. J. Chromatogr. Sci. 32:165
(1994).
42. B. Lundgren, H. Boren, A. Grimvall, and R. Savenhed, Isolation of off-flavour com-
pounds in water by chromatographic sniffing and preparative gas chromatography,

J. Chromatogr. 482:23 (1989).
43. J. P. H. Linssen, J. L. G. Janssens, J. P. Roozen, and M. A. Posthumus, Combined
gas chromatography and sniffing port analysis of volatile compounds of mineral
water packed in polyethylene laminated packages, Food Chem. 46:367 (1993).
44. R. Cantagrel and J. P. Vidal, Research on compounds responsible for cork taint in
cognacs, Flavors and Off-Flavors (G. Charalambous, ed.), Elsevier, Amsterdam,
1990, p. 139.
45. C. R. Strauss and T. Heresztyn, 2-Acetyltetrahydropyridinesa cause of the
mousy taint in wine, Chem. Ind. 109 (1984).
46. A. Rapp, P. Pretorius, and D. Kugler, Foreign and undesirable flavours in wine, Off-
flavors in Foods and Beverages (G. Charalambous, ed.), Elsevier, Amsterdam, 1992,
p. 485.
47. L. M. Seitz, R. L. Wright, R. D. Waniska, and L. W. Rooney, Contribution of 2-
acetyl-1-pyrroline to odors from wetted ground pearl millet, J. Agric. Food Chem.
41:955 (1993).
48. U. C. Konopka, H. Guth, and W. Grosch, Potent odorants formed by lipid peroxida-
tion as indicators for the warmed-over-flavor (WOF) of cooked meat, Z. Lebensm.
Unters. Forsch. 201:339 (1995).
49. R. Teranishi and S. Kint, Sample preparation, Flavor Science. Sensible Principles
and Techniques (T. E. Acree and R. Teranishi, eds.), American Chemical Society,
Washington, DC, 1993, p. 137.
50. N. Abbott, P. X. Etievant, D. Langlois, I. Lesschaeve, and S. Issanchou, Evaluation
of the representativeness of the odor of beer extracts prior to analysis by GC eluate
sniffing, J. Agric. Food Chem. 41:777 (1993).
51. P. Etievant, L. Moio, E. Guichard, D. Langlois, I. Lesschaeve, P. Schlich, and E.
Chambellant, Aroma extract dilution analysis (AEDA) and the representativeness
of the odour of food extracts, Trends in Flavour Research (H. Maarse and D. G.
van der Heij, eds.), Elsevier, Amsterdam, 1994, p. 179.
52. I. Blank, A. Stampfli, and W. Eisenreich, Analysis of food flavourings by gas chro-
matography-olfactometry, Trends in Flavour Research (H. Maarse and D. G. van
der Heij, eds.), Elsevier, Amsterdam, 1994, p. 271.
53. G. B. Nickerson and S. T. Likens, Gas chromatographic evidence for the occurrence
of hop oil components in beer, J. Chromatogr. 21:1 (1966).
54. L. Maignial, P. Pibarot, G. Bonetti, A. Chaintreau, and J. P. Marion, Simultaneous
distillation-extraction under static vacuum: isolation of volatile compounds at room
temperature, J. Chromatogr. 606:87 (1992).
55. A. Sen, G. Laskawy, P. Schieberle, and W. Grosch, Quantitative determination of
-damascenone in foods using a stable isotope dilution assay, J. Agric. Food Chem.
39:757 (1991).
56. N. Fischer and F. J. Hammerschmidt, A contribution to the analysis of fresh straw-
berry flavour, Chem. Mikrobiol. Technol. Lebensm. 14:141 (1992).
57. P. Schreier, Quantitative composition of volatile constituents in cultivated strawber-
ries, Fragaria ananassa cv. Senga Sengana, Senga Litessa and Senga Gourmella, J.
Sci. Food Agric. 31:487 (1980).
58. W. Pickenhagen, A. Velluz, J. P. Passerat, and G. Ohloff, Estimation of 2,5-
330 Blank
dimethyl-4-hydroxy-3(2H)-furanone (FURANEOL) in cultivated and wild straw-

berries, pineapples and mangoes, J. Sci. Food Agric. 32:1132 (1981).
59. P. Schieberle, Heat-induced changes in the most odour-active volatiles of strawber-
ries, Trends in Flavour Research (H. Maarse and D. G. van der Heij, eds.), Elsevier,
Amsterdam, 1994, p. 345.
60. T. Hofmann, P. Schieberle, and W. Grosch, Oxidative stability of thiols formed by
the Maillard reaction, J. Agric. Food Chem. 44:251 (1996).
61. E. Block, Flavor artefacts, J. Agric. Food Chem. 41:692 (1993).
62. E. Block, S. Naganathan, D. Putman, and S. H. Zhao, Allium chemistry: HPLC
analysis of thiosulfinates from onion, garlic, wild garlic (Ramsoms), leek, scallion,
shallot, elephant (great-headed) garlic, chive, and Chinese chive, J. Agric. Food
Chem. 40:2418 (1992).
63. E. Block, D. Putman, and S. H. Zhao, Allium chemistry: GC-MS analysis of thiosul-
finates and related compounds from onion, leek, scallion, shallot, chive, and Chinese
chive, J. Agric. Food Chem. 40:2431 (1992).
64. I. Blank, A. Sen, and W. Grosch, Potent odorants of the roasted powder and brew
of Arabica coffee, Z. Lebensm. Unters. Forsch. 195:239 (1992).
65. K. Grob Jr., G. Grob, and K. Grob, Comprehensive, standardized quality test for
glass capillary columns, J. Chromatogr. 156:1 (1978).
66. A. A. Williams and D. S. Mottram, Gas chromatographic analysis of furaneol, J.
HRC CC 4:421 (1981).
67. M. H. Boelens and L. J. van Gemert, Physiochemical parameters related to organo-
leptic properties of flavour components, Developments in Food Flavours (G. G.
Birch and M. G. Lindley, eds.), Elsevier, London, 1986, p. 23.
68. I. Blank, K. H. Fischer, and W. Grosch, Intensive neutral odorants of linden honey.
Differences from honeys of other botanical origin. Z. Lebensm. Unters. Forsch. 189:
426 (1989).
69. I. Blank and W. Grosch, Evaluation of potent odorants in dill seed and dill herb
(Anethum graveolens L.) by aroma extract dilution analysis. J. Food Sci. 56:63
(1991).
70. H. Guth, Verderb von Sojaol unter Einwirkung von Licht und SauerstoffIdentifi-
zierung der Aromastoffe und Vorlaufer, Thesis, Technical University of Munich,
Munich, 1991, p. 56.
71. L. J. van Gemert, Report of the workshop Gas chromatography-olfactometry,
Trends in Flavour Research (H. Maarse and D. G. van der Heij, eds.), Elsevier Sci-
ence, Amsterdam, 1994, p. 305.
pounds and their precursors, Characterization of Food: Emerging Methods (A. G.
Gaonkar, ed.), Elsevier Science, Amsterdam, 1995, p. 403.
73. N. Abbott, P. X. Etievant, S. Issanchou, and D. Langlois, Critical evaluation of two
commonly used techniques for the treatment of data from extract dilution sniffing
analysis, J. Agric. Food Chem. 41:1698 (1993).
74. J. E. R. Frijters, A critical analysis of the odor unit number and its use. Chem. Senses
Flav. 3:227 (1978).
75. J. R. Piggott, Relating sensory and chemical data to understand flavor, J. Sens. Stud.
4:261 (1990).
76. M. Guntert, H. J. Bertram, R. Hopp, W. Silberzahn, H. Sommer, and P. Werkhoff,

Thermal generation of flavor compounds from thiamin and various amino acids,
Recent Developments in Flavor and Fragrance Chemistry (R. Hopp and K. Mori,
eds.), VCH, Weinheim, 1993, p. 215.
77. T. E. Acree and J. Barnard, Gas chromatography-olfactometry and CharmAnal-
ysis, Trends in Flavour Research (H. Maarse and D. G. van der Heij, eds.), Else-
vier, Amsterdam, 1994, p. 211.
78. W. Grosch, Evaluation of the key odorants of food by dilution experiments, among
models and omission, Chem. Senses 26:533 (2001).
79. H. Guth and W. Grosch, Identification of the character impact odorants of stewed
beef juice by instrumental analyses and sensory studies, J. Agric. Food Chem. 42:
2862 (1994).
80. P. Semmelroch and W. Grosch, Studies on character impact odorants of coffee
brews, J. Agric. Food Chem. 44:537 (1996).
81. W. Grosch, Determination of potent odourants in foods by aroma extract dilution
9:147 (1994).
12
Quantitative Use of Gas
ChromatographyOlfactometry:
The GC-SNIF Method
Alain Chaintreau
Firmenich S.A., Geneva, Switzerland
I. INTRODUCTION
Gas chromatographyolfactometry (GC-O) has now become a widely used

method that associates the separation capability of gas chromatography to the
specific sensitivity of the human nose as a detector. It was first mentioned in the
literature in 1964 (1), after separation of volatiles in packed columns. However,
its use in flavor and fragrance laboratories might well be older (2). At that time
it was only of limited qualitative assistance to flavorists and perfumers in the
evaluation of odor constituents. It is still used in this informal manner in many
laboratories.
The pioneer of the scientific approach to GC-O was T. Acree, who rational-
ized and computerized the technique in 1984 (3). His method, called CharmAna-
lysis (combined hedonic aroma response measurement), is based on successive
dilutions and GC-injection of a flavor extract, until the assessor no longer detects
the odor at the sniffing port. For each GC-elution, the assessor presses a button
during the perception of odors, to generate individual olfactograms made of a
series of square signals. After data treatment, a computer-generated global olfac-
togram assigns greater importance to odor peaks that are smelt in the highest
dilution of the extract.
Three years later, W. Grosch proposed a simpler techniquethe aroma
extract dilution analysis (AEDA)based on a similar dilution approach (4). For
each odor peak of the olfactogram, the highest dilution level of the extract at
333
334 Chaintreau
which it is still perceivable is noted. The olfactogram is constituted of bars re-

porting the dilution factor (FD) at which odorants become undetectable. The
method is easier to use than CharmAnalysis because no computer is required,
but the continuous signal recording over the whole GC run duration is lost.
A third technique, OSME (smell, in Greek), was proposed in 1989 by Mc-
Daniel to follow the odor intensities of peaks eluting from the GC column, by
moving a cursor over the olfactogram duration (5,6). It requires only the injection
of the aroma extract at one level of dilution, and a time-dependant signal is contin-
uously recorded by the computer.
These three techniques have allowed the determination of the most impor-
tant aroma compounds in many foods. Both dilution techniques, CharmAnalysis
and AEDA, have been reviewed (712). Applications of OSME do not seem to
have been the subject of any specific review, but the technique is referenced by
Mistry (13).
Until recently, GC-O was applied mainly for screening key odorants. How-
ever, there is an increasing need for more and more quantitative applications,
such as olfactogram comparisons or relating the perception of an odorant to its
concentration within the aroma. The first case requires evaluating the significance
of a difference between two odor measurements. The second implies the estab-
lishment of a calibration curve.
In this chapter, olfactogram is preferred to aromagram because the
former is more general and the latter is restricted to a food odor. The theoretical
justifications of GC-SNIF will not be discussed here [see (14)]. For techniques
of aroma sampling and qualitative applications of GC-O, the reader can refer to
Chapter 11. The present chapter focuses on quantitative aspects of GC-O that
can be solved using GC-SNIF. To draw conclusions about its capabilities,
this chapter makes a synthesis from recent studies.
II. QUANTITATIVE ABILITY OF PREVIOUSLY EXISTING

TECHNIQUES
A. AEDA
A critical evaluation of dilution methods (15) has revealed the existence of gaps
in coincident responses. In other terms, an odorant can be undetected by a panel-
ist at a given dilution level, but it may be perceived by the same panelist at higher
dilutions. W. Grosch attributed these gaps to inexperience (7), but recent studies
do not support this observation: Hanaoka proved that the panelists breathing rate
was statistically correlated with the risk of missing odorants, due to the interrup-
tion of smelling, when breathing out (16). In addition, systematic omissions may
occur, as observed by Priser, who used AEDA to compare impact aroma com-
pounds of three different Champagne wines (17). The second- and third-most
The GC-SNIF Method 335
FIGURE 1 The three highest FD values in Champagne wine H, according to the first
panelist (gray). The one recorded values for the second panelist (black). Values for the
same odorants in wines M and J. (Adapted from Ref. 17.)
important odorants detected by the first assessor in wine H were not perceived
at all by the second one (Fig. 1). The same two odorants remained undetected by
the same panelist in two other wine extracts, M and J. This reveals the existence of
specific anosmias inherent to human beings that cannot be overcome by any train-
ing or experience.
The ratio of the concentration of an odorant to its odor threshold gives the
odor activity value (OAV). This is often used to rank the olfactory importance
of aroma constituents. However, the OAV concept has been criticized (18,19)
because it assumes that a linear relationship exists between the odor intensity
and the odorant concentration. This is in conflict with most accepted psychophysi-
cal laws based on logarithmic or exponential relationships (20). A recent article
by Reineccius and coworkers confirms that OAVs do not represent the contribu-
tion of odorants in the overall aroma and that using OAVs to rank their contribu-
tions could result in a misinterpretation (21).
According to its inventor and main users (7,11), AEDA is designed only for
the screening of impact odorants. Its advantage (simplicity of use) also appears as
a limitation to the development of a real quantitative tool because it is hardly
applicable to more than one or two assessors.
B. CharmAnalysis
As CharmAnalysis is a dilution technique, gaps in coincident responses and
interindividual differences in sensitivity may also affect the olfactogram gener-
ated by a single panelist. To evaluate this variability, Acree submitted replicates
336 Chaintreau
of a CharmAnalysis to a statistical analysis (22). He evaluated the variance and

the least significant difference between CharmAnalysis peaks. In agreement with
these results, Guichard et al. concluded the need for three GC-O replications with
10 dilution levels for a representative evaluation of peak intensities (15). It was
suggested that a quantitative use of CharmAnalysis is possible, but the replication
of GC-O run would take weeks.
To obtain responses closer to the perceived odor intensities, Friedrich and
Acree proposed the transformation of the CharmAnalysis or OAV into an odor
spectrum using Stevens law (9,10). However, a median value of 0.5 was used
as a Stevens exponent, whereas this number may vary significantly from one
compound to another [e.g., 0.24 for guaiacol (23)].
C. OSME
As originally published, the direct measurement of peak intensities did not appear
to be very reproducible. Large differences were observed between individuals,
or within replications of the same panelist (5). Simultaneously to the investigation
of CharmAnalysis variability, the standard deviation and the least significant dif-
ference between two OSME-peaks were calculated for each of the four panelists.
However, the relationship between the subject response and the stimulus (concen-
tration) remained unclear (6). The technique was improved by using finger span
measurements to evaluate odor intensities (15). From these trials performed by
10 panelists, reproducibility was calculated by grouping panelists in two teams
of five persons (14), leading to a mean relative standard deviation of about 30%,
with extremes up to 126%. New improvements recently reported by Etievant
et al. (23) show a log/log relationship between the mean peak height based on
the finger span measurement of four trained panelists and the concentration of
the odorant. From these last developments, OSME seems to be a promising tech-
nique for quantitative purposes. However, it requires a significant amount of time
to initially train the panelists.
III. THE GC-SNIF METHOD
The GC-SNIF method has been developed to achieve the following criteria:
Quickness. No necessity to train panelists or to perform a series of dilutions.
Simplicity. Panelists must focus only on the odor detection and generate
an instinctive response without being disturbed by giving, for example,
intensities.
Reproducibility. The lack of reproducibility between individuals exempli-
fied in all methods implied the use of a panel.
Easy generation of an olfactogram. This was achievable using continuous

recording of a signal generated by each panelist, similar to CharmAna-
lysis and OSME, and a computerized treatment of individual olfacto-
grams.
A. Principle of the Method

During a GC-O acquisition, each panelist continuously smells odors eluting from
the chromatographic column and presses a button for the whole duration of the
perception of a given odorant. This operation generates, on the PC screen, a series
of square signals called the olfactogram (Fig. 2).
After GC-O detection is repeated with the different members of the panel,
the resulting individual olfactograms are averaged (Fig. 2). Each coincident re-
sponse of panelists gives a signal, whose height represents the number of panelists
having detected an odor at this retention time. After normalization of the mean
olfactogram to 100% (100% peak detected by all panelists), the resulting peak
height indicates the detection frequency of this odorant by the panel. Therefore
the peak height and its area have been respectively called:
NIF: Nasal impact frequency
SNIF: Surface of nasal impact frequency
FIGURE 2 Principle of the GC-SNIF data treatment. (Reprinted with permission from
Ref. 14, copyright 1997, ACS.)
338 Chaintreau
The statistical analysis of individual olfactograms generated by a large panel has

shown an optimal composition of between six and ten members for routine analy-
ses with a NIF uncertainty of 20% and 5%, respectively (14).
In case of a too-long GC duration (more than 25 minutes), panelist attention
may decrease. To avoid fatigue, it is preferable that two persons consecutively
participate in the generation of an individual olfactogram. In a second GC run,
the same, two panelists will smell each half run in the reverse order. As individual
olfactograms are finally averaged, the risk of missing an odor is low, as long as
the change of panelist is carried out at a different retention time in each individual
olfactogram.
B. Screening of Impact Aroma Compounds

Because the GC-SNIF analysis has been designed for quantitative pur-
poses, it is obviously applicable to the qualitative determination of impact odor-
ants of a given scent or aroma. It even lowers the risk of missing important odor
contributors inherent in the use of methods based on only one or two panelists
(Fig. 1).
In 1993, before the quantitative ability of GC-SNIF was established
(24,25), it was first applied to the qualitative determination of Capsicum annuum
key odorants (26) and of the off-odors in mineral water (27), using a panel of
10 to 12 persons. It has also been used to find out the aroma impact components
of French beans (24,28), coffee (25,29), yogurt (25,30), vinegar (31), mussels
(32), pickles (33), and wines of Champagne (17,34), Bourgogne (35), and Alsace
(36). The moldy/earthy defect of green coffee has been elucidated using the same
technique (37). To our knowledge, there are no reports on the application of GC-
SNIF to the field of perfumery.
C. Quantitative Olfactogram Comparisons

Comparing two olfactograms requires the determination of two parameters:
1. The standard deviation of the NIF or SNIF values, to evaluate the vari-
ability of an olfactogram from one experiment to another
2. The least significant difference (LSD) between NIF (or SNIF) values
of a same compound smelt in the olfactograms of two products
Using (a) a model mixture of volatiles and (b) a real product (pet feces), the
mean relative standard deviation (RSD) has been calculated from olfactograms
generated by the same panel or by two different panels (Table 1) (14). Mean
RSDs of NIFs and SNIFs were in the 13.818% range, for a six-member panel.
TABLE 1 Repeatability, Reproducibility, and Least Significant Difference

of/between GC-O Peaks
Least
significant
Relative difference
standard (95%
deviation confidence)
Number of Time NIF SNIF

panelists interval (%) (a) NIF SNIF Reference
Cocktail (repeatabil- 6 1 wk 14.1% 13.8% 31% 2085 (14)

ity)
Cocktail (reproduc- 6 1 wk 15.7% 14.8% 33% 2639 (14)
ibility)
Pet excrement 6 1 wk 17% 18% 35% 2351 (14)
(repeatability)
1-Octen-3-one a 8 4 mo 20.2% 26.5% n.d. n.d. (38)
(repeatability)
1-Octen-3-one a 12 4 mo 12.9% 1.9% n.d. n.d. (38)
(repeatability)
a
In a coffee brew.
n.d. not determined.
This corresponds to variations of not more than one panelist of each olfactometric
peak between replications of the olfactogram. Over a 4-month period, the repeat-
ability remained within a one-panelist variation (Table 1).
From these RSD values, LSDs were obtained using the same panel, as well
as by using two different panels (Table 1). LSDs varied in the range of 2535
NIF%, or 16872639 SNIF units, depending on the level of confidence. Similar
values were obtained by Aubry (31% NIF ) (35) and Bernet (33% NIF ) (36),
using panels of 16 and 18 persons, respectively. However, the nondependence
of LSD versus the panel size should be re-investigated, because the three pub-
lished values were calculated with three different tests, and Bernets LSD evalua-
tion was based on three replications by six judges, instead of 18 independent
persons.
It must be pointed out that obtaining similar RSDs and LSDs using two
different panels as repeating GC-O analyses with the same one indicates that two
independent panels were able to generate similar olfactograms (14).
As an increase of NIF or SNIF greater than the LSD indicates a concentra-
tion increase of the corresponding odorant (14), this allows the quantitative com-
parison of olfactograms. Such a capability has been applied to real aromas. Based
340 Chaintreau
FIGURE 3 Aroma comparison of a plain yogurt with the corresponding milk. (Adapted
from Ref. 30.) Marked peaks indicate the odorants generated by the fermentation.
on the SNIF increase of impact odorants from milk to yogurt, Ott et al. determined
the aroma contributors generated by the fermentation (Fig. 3) (30). SNIF varia-
tions below the LSD level were considered to be nonsignificant. In a later paper,
the quantitation of the corresponding compounds confirmed the direction of the
SNIF variation, even in case of a lower significance (Table 2) (39).
TABLE 2 SNIF and Concentration Variations of Impact Odorants from

Milk to Yogurt
Ret. Variation Variation
index Compound (SNIF) Significance a (ppm/%)
716 Acetaldehyde 7006 16,6/
757 Dimethylsulfide 1300 0.005/20%
993-5 2,3-Butanedione 5492 1.35/
1082 2,3-Pentanedione 3549 0.13/
2002 Benzothiazole 1707 0.09/24%
a
Significance of the NIF variation at 95% confidence using a same panel.
Source: Adapted from Ref. 39.
The GC-SNIF method has also allowed the elucidation of the process-
induced modifications of the coffee aroma by comparing impact odorants of an
instant coffee with the corresponding brew (29). It was applied as well to the
characterization of three Champagne wines (17), and to determine which were
the impact odorants protected by the addition of an antioxidant in the wine (34).
Because products that can be smelled can be differentiated by their GC-
SNIF olfactograms, it becomes possible to classify samples using statistical
tools. In the case of antioxidant addition in Champagne wines, a factorial corre-
spondence analysis was performed (Fig. 4) (34). The first axis (F1) differentiates
wines that have been protected with an antioxidant (BHA), and the third axis (F3)
distinguishes between years of wine production (Ch1 and Ch2). BHA-containing
wines were characterized by a higher odor impact of diethyl succinate, ethyl
cinnamate, and -undecalactone.
A similar approach has been used to classify four wines of Bourgogne
according to their impact odorants (35).
D. GC-O Quantitation
The first quantitative application of the relationship between the detection fre-
quency and concentration of an odorant was made by van Ruth et al. (24), who
FIGURE 4 Factorial correspondence analysis of impact odorants characterizing two

Champagne wines. Only the first (F1) and third axes (F3) are shown. (Reprinted with
permission from Ref. 34, copyright 1999, ACS.)
342 Chaintreau
found that the first parameter varied linearly with the logarithm of the second.
This observation was used to monitor the release rate of impact odorants from
rehydrated French beans.
According to the theoretical basis of GC-SNIF, the NIF function versus
the concentration logarithm is a sigmoid. This is not contradictory to van Ruths
observation, as the central part of a sigmoid well approximates to a straight line
(40). However, working with a sigmoid is tedious, and its linearization using
Probits makes the calibration easier (Fig. 5) (14,38). As the study of the NIF
variability was shown to be equal to or less than one panelist contribution over
a panel of 6 to 12 members, this variation can be transformed into a confidence
interval around the Probit calibration line (Fig. 5).
Until recently, there has been only one report of an attempt to quantitate
an odorant using GC-O. Probits of NIF values of 1-octen-3-one in a model solu-
tion and in coffee were compared to a calibration curve (38). Results showed
that GC-SNIF can compete with most sensitive and selective techniques, such
as tandem-MS, to quantify extremely intense odorants. In the present example,
the GC-O sensitivity was 75 to 500 times higher than MS for the quantitation
procedure.
FIGURE 5 GC-SNIF calibration curve and confidence interval of 1-octen-3-one for

eight and 12 panelists. (Reprinted with permission from Ref. 38, copyright 1999, ACS.)
IV. COMPARISON WITH OTHER METHODS
From a qualitative point of view, the main GC-O methods seem to be equivalent
for determining the impact odorants of a product, with the exception that some
peaks can be missed when using only one or two panelists (Fig. 1). The aroma
impact compounds of coffee brew found by GC-SNIF (29) were in agreement
with those found by AEDA (41). Le Guen et al. compared OSME, AEDA, and
GC-SNIF results to determine the most potent odorants of cooked mussels
(32). They concluded that the three methods were well correlated. They also
observed that GC-SNIF was twice as fast as AEDA and OSME.
The greater rapidity of the GC-SNIF over AEDA and OSME was also
noted by Priser, who additionally found that OSME and GC-SNIF gave simi-
lar profiles from a more quantitative point of view (17). Starting from her normal-
ized scores that were determined for each attribute using the three techniques,
we have recalculated the euclidian distance between the three sensory profiles
(Fig. 6). The overall distance between OSME and GC-SNIF results is always
closer than that between OSME and AEDA, or GC-SNIF and AEDA.
The proximity of results obtained by OSME and GC-SNIF might be
justified by van Ruths (42) observation that the number of assessors perceiving
an odour correlated significantly with odor intensity scores [. . .] indicating that
the number of assessors is a sufficient measure of the odour intensity. When
comparing NIF values and odor intensities of all representative odorants per-
ceived in three different wines, Bernet also found a good linear relationship be-
tween both variables (Fig. 7) (36).
Up to now, a real comparison of quantitative performances between GC-
SNIF and other GC-O techniques has not yet been published.
FIGURE 6 Euclidian distance between the sensory profiles obtained with the three
GC-O techniques for three different Champagne wines (H, M, J). (Adapted from Ref. 36.)
344 Chaintreau
FIGURE 7 Comparison of NIFs and odor intensity responses for all representative odors
detected in three different Gewurztraminer wines. (Adapted from Ref. 36.)
V. LIMITATIONS AND FUTURE IMPROVEMENTS

A. Sample Preparation
As originally designed, GC-SNIF includes the headspace sampling of the
product to be analyzed. Of course, the GC-SNIF data treatment is also applica-
ble to any other injection mode, such as a liquid extract. However, the headspace
(HS) composition is the most representative of the odor perceived by the nose,
as long as the method used to collect volatiles does not alter their relative
amounts. The static-and-trapped headspace technique (S&T-HS) does not distort
the HS composition as a function of volatilities, and allows the collection of a
greater vapor volume than a gas syringe (14,43). However, usual HS trapping
agents still have some limitations (limited breakthrough volumes for very volatile
compounds, alteration of very labile compounds upon desorption, etc.). Cold
traps are prone to be blocked by the formation of ice. Therefore, extending the
performances of HS trapping agents would be a great help for any GC-O tech-
nique.
B. Chromatographic Problems
When performing the evaluation of odorants after GC separation, odors are smelt
out of their context, i.e., in an isolated form. Up to now, little is known about their
possible synergies and/or their relative contributions between odorants perceived

simultaneously (21). Attempts to ascertain GC-O results by reformulating a flavor
with these impact odorants are rare [e.g., (12,33,44)] and not always conclusive
(45). Conversely, peak coelutions may occur with all techniques, giving rise to
the overevaluation of the impact of a given odorant.
The injection into a GC column of a series of components in the same
amount is known to give different peak heights and peak widths as a function
of their retention time. Does it affect their odor intensities perceived at the sniffing
port (18)? This has never been investigated, in spite of the fact that it might
significantly change their real odor impact and modify their contribution ranking
in the overall flavor.
All three phenomena are inherent in all GC-O techniques. Therefore, con-
clusions drawn from the character of isolated constituents using such a technique
should be applied with care to interpret the character of the overall odor.
C. Analysis Duration
Although GC-SNIF seems to perform better than other techniques (17,32),
in terms of data acquisition and data treatment duration, obtaining a final olfacto-
gram still takes days, versus weeks for dilution methods with several judges
(13,15,22). To enable analysts to routinely perform GC-O runs with quantita-
tively reliable results, the analysis duration should be still decreased.
VI. CONCLUSION
From the literature mentioned in this chapter, the GC-SNIF technique already
exhibits very unique capabilities in terms of quickness and reproducibility and
it opens the quantitation field to the GC-O analysis. The superiority of a technique
involving a panel instead of one or two assessors is now well established. How-
ever, the exact limits are not well known, as they are still under investigation.
For instance, it is still unclear in which kind of applications the GC-SNIF or
OSME method is more appropriate. As GC-O is the link between chemical and
sensorial analyses, it is important to flavor and fragrance research, and improve-
ments in GC-O techniques will continue as the technology evolves and matures.
ACKNOWLEDGMENTS
We gratefully acknowledge Mrs. R. Naef, Mr. M. Baumgartner, and Dr. A. Clark

for the revision of this manuscript. We are also grateful to Drs C. Priser, N.
346 Chaintreau
Dirninger, and P. Etievant for providing us with some raw data and figures used
in this paper.
REFERENCES
1. G. H. Fuller, R. Steltenkamp, and G. A. Tisserand, The gas chromatograph with

human sensor: perfumer model, Ann. NY Acad. Sci. 116:711 (1964).
2. L. Peyron, personal communication, and: Les parfumeurs et la chromatographie,
La France et ses parfums 10:199 (1967).
3. T. E. Acree, J. Barnard, and D. G. Cunningham, The analysis of odor-active volatiles
in gas chromatographic effluents, Analysis of Volatiles, de Gruyter, New York,
1984.
4. F. Ullrich and W. Grosch, Identification of most intense volatile flavour compounds
formed during autooxidation of linoleic acid, Z. Lebensm. Unters Forsch. 184:277
(1987).
5. M. R. McDaniel, R. Miranda-Lopez, B. T. Watson, N. J. Michaels, and L. M. Libbey,
Pinot noir: a sensory/gas chromatographic approach, Flavors and Off-Flavors (G.
Charalambous, ed.), Elsevier Science B. V., Amsterdam, 1989.
6. M. A. A. P. Da Silva, D. S. Lundahl, and M. R. McDaniel, The capability and
psychophysics of OSME: A new GC-olfactometry technique, Trends in Flavour Re-
search (H. Maarse and D. G. van der Heij, eds.), Elsevier Science B.V., Amsterdam,
1994.
7. W. Grosch, Determination of potent odourants in foods by aroma extract dilution
9:147 (1994).
8. T. Acree, Bioassay for flavor, Flavor Science. Sensible Principles and Techniques
(T. Acree and R. Teranishi, eds.), ACS, Washington, DC, 1993.
9. Y. G. Feng and T. E. Acree, Gas chromatography olfactometry in aroma analysis,
Foods Food Ingredients J. Jpn. 179:57 (1999).
10. J. E. Friedrich and T. Acree, Gas chromatography olfactometry (GC/O) of dairy
products, Int. Dairy Journal 8:235 (1998).
11. I. Blank, Gas chromatography-olfactometry in food aroma analysis, Techniques for
Analyzing Food Aroma (R. Marsili, ed.), Marcel Dekker, New York, 1996.
pounds and their precursors, Characterization of Food: Emerging Methods (A. G.
Gaonkar, ed.), Elsevier, Amsterdam, 1995.
13. B. S. Mistry, T. Reineccius, and L. K. Olson, Gas Chromatography-olfactometry
Aroma (R. Marsili, ed.), Marcel Dekker, New York, 1997.
14. P. Pollien, A. Ott, F. Montigon, M. Baumgartner, R. Munoz-Box, and A. Chain-
treau, Hyphenated headspace-gas chromatography-sniffing: screening of impact
odorants and quantitative aromagram comparisons, J. Agric. Food Chem. 45:2630
(1997).
15. H. Guichard, E. Guichard, D. Langlois, S. Issanchou, and N. Abbott, GC Sniffing
analysis: olfactive intensity measurement by two methods, Z. Lebensm. Unters

Forsch. 201:344 (1995).
16. K. Hanaoka, N. Vallet, P. Giampaoli, B. Heyd, and P. MacLeod, Possible influence
of breathing on detection frequency and intensity rating in gas chromatography-
olfactometry, Food Chem. 72:97 (2000).
17. C. Priser, Contribution a la characterisation objective de larome de vin de Cham-
pagne par des methodes sensorielles et physico-chimiques, Thesis (Universite de
Bourgogne), Dijon, 1997.
18. F. Drawert and N. Christoph, Significance of the sniffing-technique for the determi-
nation of odor thresholds and detection of aroma impacts of trace volatiles, Analysis
of Volatiles: Methods and Applications (P. Schreier, ed.), De Gruyter, Berlin, 1984.
19. J. E. R. Frijters, Some psychophysical notes on the use of the odour unit number,
Progress in Flavour Research (D. G. Land and H. E. Nursten, eds.), Applied Science,
London, 1979.
20. Evaluation sensorielle. Manuel methodologique, 2nd ed., Lavoisier, Paris (1998).
21. V. Audouin, F. Bonnet, Z. M. Vickers, and G. Reineccius, Limitations in the use of
odor activity values to determine important odorants in food, Gas Chromatography-
Olfactometry: the State of the Art (J. V. LeLand, P. Schieberle and A. J. Taylor,
eds.), ACS, 2001.
22. T. Acree and J. Barnard, Gas chromatography-olfactometry and CharmAnalysis,
Trends in Flavour Research (H. Maarse and D. G. van der Heij, eds.), Elsevier Sci-
ence B.V., Amsterdam, 1994.
23. P. Etievant, G. Callement, D. Langlois, S. Issanchou, and N. Coquibus, Odor inten-
sity evaluation in gas chromatography-olfactometry by finger span method, J. Agric.
Food Chem. 47:1673 (1999).
24. S. M. Van Ruth, J. P. Roozen, and J. L. Cozijnsen, Gas chromatography/sniffing
port analysis evaluated for aroma release from rehydrated french beans (Phaesolus
vulgaris), Food Chem. 56:343 (1996).
25. A. Ott, P. Pollien, F. Montigon, and A. Chaintreau, Improved headspace-GC-sniffing
technique: screening of aroma impact flavorings. 4th Int. Symp. Hyphenated Tech-
niques in Chromatography. Bruge (B). February 79, 1996.
26. S. M. Van Ruth, J. P. Roozen, and J. L. Cozijnsen, Changes in flavour release from
rehydrated diced bell peppers (Capsicum annuum) by artificial saliva components
in three mouth model systems, J. Sci. Food Agric. 67:189 (1995).
27. J. H. Linssen, J. L. G. M. Janssen, J. P. Roozen, and M. A. Posthumus, Combined
gas chromatography and sniffing port analysis of volatile compounds of mineral
water packed in polyethylene laminated packages, Food Chem. 46:367 (1993).
28. S. M. Van Ruth, J. P. Roozen, D. F. Nahon, J. L. Cozijnsen, and M. A. Posthumus,
Flavour release from rehydrated french beans (Phaesolus vulgaris) influenced by
composition and volume of artificial saliva, Z. Lebensm. Unters Forsch. 203:1
(1996).
29. P. Pollien, Y. Krebs, and A. Chaintreau, Comparison of a brew and an instant coffee
using a new GC-olfactometric method, Association Scientifique Internationale du
Cafe 1997:191 (1998).
30. A. Ott, L. B. Fay, and A. Chaintreau, Determination and origin of the aroma impact
compounds of yogurt flavor, J. Agric. Food Chem. 45:850 (1997).
348 Chaintreau
31. M. Charles, B. Martin, C. Ginies, P. X. Etievant, G. Coste, and E. Guichard, Potent

aroma compounds of two red wine vinegars, J. Agric. Food Chem. 48:70 (2000).
32. S. Le Guen, C. Prost, and M. Demainay, Critical comparison of three olfactometric
methods for the identification of the most potent odorants in cooked mussels (Mytilus
edulis), J. Agric. Food Chem. 48:1307 (2000).
33. R. Marsili and N. Miller, Determination of major aroma impact compounds in fer-
mented cucumbers by solid-phase microextraction-gas chromatography-mass spec-
trometry-olfactometry detection, J. Chromatogr. Sci. 38:307 (2000).
34. A. Escudero and P. Etievant, Effect of antioxidants on the flavor characteristics and
the gas chromatography/olfactometry profiles of Champagne extracts, J. Agric. Food
Chem. 47:3308 (1999).
35. V. Aubry, Contribution a la connaissance aromatique des vins de pinot noir de Bour-
gogne par des methodes instrumentales et sensorielles. Etude de la potentialite olfac-
tive de quatre composes darome, Thesis (Universite de Bourgogne), Dijon, 1999.
36. C. Bernet, Contribution a la connaissance des composes darome cles des vins du
cepage Gewurztraminer cultive en Alsace, Thesis (Universite de Bourgogne), Dijon,
2000.
37. E. Cantergiani, H. Brevard, R. Amado, Y. Krebs, A. M. Feria-Morales, and C. Yeret-
zian, Characterisation of mouldy/earthy defect in green Mexican coffee, Association
Scientifique Internationale du Cafe, 18th Coll. 43 (2000).
38. P. Pollien, L. B. Fay, M. Baumgartner, and A. Chaintreau, First attempt of odorant
quantitation using gas chromatography-olfactometry, Anal. Chem. 71:5391 (1999).
39. A. Ott, M. Baumgartner, and A. Chaintreau, Aroma comparisons of traditional and
mild yogurt: headspace-GC quantitation of volatiles and origin of -diketones, J.
Agric. Food Chem. 47:2379 (1998).
40. A. Chaintreau and P. Pollien. Aspects quantitatifs de lanalyse GC-sniffing par la
methode GC-SNIF. Rivista Italiana EPPOS, Gennaio:258 (1999).
41. P. Semmelroch and W. Grosch, Analysis of roasted coffee powders and brews by
gas chromatography-olfactometry of headspace samples, Lebensm. Wiss. Technol.
28:310 (1999).
42. S. M. Van Ruth, J. P. Roozen, M. E. Hollmann, and M. A. Posthumus, Instrumental
and sensory analysis of the flavour of French beans (Phaesolus vulgaris) after differ-
ent rehydration conditions, Z. Lebensm. Unters Forsch. 203:7 (1996).
43. A. Chaintreau, A. Grade, and R. Munoz-Box, Determination of partition coefficients
and quantitation of headspace volatile compounds, Anal. Chem. 67:3300 (1995).
44. H. Guth and W. Grosch, Identification of the character impact odorants of stewed
beef juice by instrumental analyses and sensory studies, J. Agric. Food Chem. 42:
2862 (1994).
45. A. Plotto, J. P. Mattheis, D. S. Lundahl, and M. R. McDaniel, Validation of gas
chromatography olfactometry results for Gala apples by evaluation of aroma-
active compounds mixtures, Flavor Analysis (C. J. Mussinan and M. J. Morello,
eds.), ACS, Washington, DC, 1998.
13
Combining Mass Spectrometry
and Multivariate Analysis to Make
a Reliable and Versatile
Electronic Nose
Ray Marsili
Dean Foods Company, Rockford, Illinois
I. INTRODUCTION
Unfortunately, the two basic techniques used to assess flavor/fragrance quality

sensory analysis and conventional GC/MSare generally too time-consuming,
complex, and labor-intensive for routine quality control application. In fact, many
of the test procedures and sample preparation methods described in this book
are inappropriate for routine quality control testing. In industrial quality control
applications, the need for speed and the large number of samples to be tested
significantly impact the type of testing procedures and instrumentation that can
be used.
Monitoring the flavor quality from raw materials to finished product is of
great importance to the food industry. For example, raw materials should be mon-
itored to ensure they have typical flavor quality, are free from taints, and that
no deterioration or contamination with off-flavor chemicals has occurred during
transport. Also, packaging materials should be checked for residual solvents to
ensure they wont contribute off-flavors to the finished product. Occasionally
packaging materials are not adequately cured prior to use, and small amounts of
solvent associated with the manufacturing of the packaging materials remain.
These residual solvents can migrate into food products, imparting malodors and
off-flavors (1,2). Monitoring flavor/off-flavor development during various pro-
349
350 Marsili
cessing steps should be conducted to ensure processes are operating correctly.

For example, unusually high levels of heat treatment during processing can cause
thermal degradation reactions to occur in a food system, resulting in the develop-
ment of a variety of off-flavors. Finished products should also be monitored to
ensure that no off-flavors have developed. Exposure to light, heat, pro-oxidant
metals, or oxygen can degrade unsaturated fatty acids in fat triglycerides and
produce foul-tasting hexanal and other aldehydes (3). Finally, monitoring finished
products over normal shelf life is a good idea. One potential problem to check
is scalpingthe absorption of desirable flavor compounds from the food product
into the packaging material (4). This type of testing is often overlooked because
of the lack of availability of suitable testing technology.
One analytical tool that has been proposed in recent years to address the
need for routine quality testing in the food industry is the electronic nose (e-nose).
An e-nose instrument has an array of weakly specific chemical sensors. Unlike
most existing chemical sensors, which are designed to detect specific chemicals,
sensors in an electronic nose are not specific to any one chemical (5).
E-nose instruments are designed to mimic the human olfactory system.
Humans have a large number of olfactory sensors in the nosetypically, approxi-
mately 10 4. Although human olfactory sensors are not specific, they are very
sensitive to certain odors. A human uses the nose to smell the aroma of a food
sample and the brain to analyze the data contributed by all the sensors in the
nose. E-noses have this same basic type of systemsensors, a method of con-
verting the sensor signals, and a component for analyzing these signals. There
are four basic components to an e-nose instrument: a mechanism to deliver the
flavor/aroma volatiles to the sensors; a sensor array; a way of converting sensor
signals to data that can be analyzed by a computer; and software analysis of
output from the sensors. Of these components, perhaps the most critical is the
sensor array.
Solid-state e-noses first appeared in the mid-1980s and have been used in
the food industry to monitor the production of beer, wine, coffee, bread, and so
forth and as a screening tool for raw materials (69). E-nose instruments are
good at assessing the chemical integrity of a samplei.e., determining whether
the sample is the same as or different than expected. However, they are not useful
for the identification and quantification of the individual components in the sam-
ple. E-noses are useful tools for answering the following types of questions:
Is our product the same today as it was yesterday?
Do we have a packaging off-flavor problem?
Is this ingredient authentic or adulterated?
Has this product changed its flavor profile during shelf life?
E-noses are most often used to provide comparative rather than quantitative
information. Because data interpretation is eliminated, the technique is ideally
Development of the Electronic Nose 351
suited for rapid QC/QA checking. The task of an e-nose instrument is similar to
that of the human in that its goal is not to determine the exact composition of
the vapor. For example, the human smells fresh and stale potato chips and assigns
them into the two appropriate categories. In reality, the two samples may contain
significantly different chemicals and/or some of the same chemicals at different
concentration levels.
Despite some promising and impressive successes, solid-state sensor-based
systems have not generally lived up to expectations. Problems with drift (short
and long term), noise, instability due to water vapor, sensor poisoning, the need
for time-consuming recalibration, poor sensor-to-sensor and instrument-to-instru-
ment reproducibility, and high instrument costs have plagued e-nose instruments
that employ conducting polymer sensors, metal oxide sensors, surface acoustic
wave devices, and other types of solid-state sensors. Because of these problems,
several manufacturers of e-nose instruments have gone out of business, and those
that continue to manufacture and market e-nose instruments have devoted much
of their research efforts to overcoming these deficiencies.
Recently, interest in using a mass spectrometry detector as an e-nose sensor
array has grown. The MS sensor approach offers important advantages over cur-
rent solid-state sensors, including no problem with water, alcohols, or poisoning;
a linear response to vapor concentrations; much less drift; and significantly im-
proved reproducibility.
There are currently over a dozen companies marketing e-nose instruments,
with most employing solid-state sensor technology rather than MS. At the time of
publication, the leading manufacturer of MS-based e-nose instruments is Agilent
Technologies (Wilmington, DE).
II. THE AGILENT 4440 CHEMICAL SENSOR

A. Operating Principle
The Agilent 4440 chemical sensor is designed as a rapid-screening tool for a
number of applications in the food, pharmaceutical, and chemical industries. The
instrument uses a headspace (HS) analyzer to transport vapors from the sample
to the quadrupole mass spectrometer. It operates by detecting fragment ions in
the range of m/z 25 to m/z 150 from volatiles present in the headspace. The data
are processed using a variety of multivariate statistical techniques. The advan-
tages of this system over conventional HS/GC or HS/GC/MS systems are its
speed, ease of use, and ability to provide answers rather than data that would
require further interpretation. Samples can be analyzed every 25 minutes, de-
pending on the amount of headspace equilibration time required. Instrument con-
trol and data analysis are so simple that operator training for predictive analysis
takes less than an hour.
352 Marsili
FIGURE 1 Operating principle of the Agilent 4440 Chemical Sensor.
The principle of the Agilent 4440 Chemical Sensor is illustrated in Fig. 1.

The equilibrated headspace gases above a sample are transferred through a capil-
lary column to the MSD. The column serves primarily as a restrictor and a flow
path from the GC inlet to the MSD. Because speed of analysis is desired, the
conventional GC capillary column can be replaced with a short uncoated deacti-
vated retention gap. There are two main advantages to using a retention gap
instead of a column: (1) it does not retain the analytes, so the sample elutes
rapidly and sample turnaround time can be as short as 2 minutes and (2) the
retention gap has no phase, so there is no column bleed, which helps to keep the
MSD source clean and background noise low.
The chemical sensor generates a mass spectrum that represents a summa-
tion of all the mass spectra of the volatiles represented in the headspace aliquot
injected. The MS data is outputted to a spreadsheet for further processing by the
Pirouette multivariate analysis software (Infometrix , Inc., Woodinville, WA).
An example of a Pirouette spreadsheet is shown in Fig. 2. The sample name
appears in the first column. The remaining columns contain mass intensity data.
The column head signifies the m/z value for the column. Each m/z (in this case,
m/z 35 to 38 are represented) would be equivalent to a chemical sensor. If masses
35 to 150 were scanned, for example, this would correspond to an array of 116
chemical sensors.
Once the data are converted to a spreadsheet format, the Pirouette software
is used to decipher meaningful trends in the MS output. One normally starts
multivariate analysis of data with an exploratory technique. With the Pirouette
FIGURE 2 Pirouette spreadsheet of mass intensity data generated by the Agilent 4440
chemical sensor.
software, two exploratory options are available: Hierarchical Cluster Analysis

(HCA) and Principal Component Analysis (PCA).
With HCA, similarities between samples (or variables) are indicated in a
graphical format called a dendrogram, a tree-shaped map constructed from the
table of distances. Branch lengths are proportional to the distances between linked
clusters. A dendrogram with 75 samples is shown in Fig. 3. The branches on the
far left of the dendrogram, called leaves, represent single samples. The length
of the branches linking two clusters is related to their similarity. The longer the
branch, the less the similarity; the shorter the branch, the greater the similarity
and, therefore, the smaller the intercluster distance. The dotted vertical line slices
through the dendrogram in Fig. 3 at a similarity value of approximately 0.5,
where four clusters can be distinguished. The lengths of these four branches are
long compared to the branches connecting clusters to the left. In random data
(i.e., dissimilar samples), leaves are often as long as or longer than most of the
branches.
A more popular exploratory multivariate analysis technique is Principal
Component Analysis (PCA). PCA is a powerful visualization tool and thus finds
use in exploratory analysis. Like HCA, it can represent graphically intersample
and intervariable relationships. Moreover, it provides a way to reduce the dimen-
354 Marsili
FIGURE 3 A dendrogram of 75 samples, depicting leaves, branches, and clusters. Dashed

line indicates a similarity value of 0.5, where four distinct clusters can be distinguished.
sionality of the data. PCA finds linear combinations of the original independent
variables (mass intensities) that account for maximal amounts of variation. An
example of a PCA plot constructed from mass intensity data appears in Fig. 4.
This plot shows Reference (good flavor quality) raw materials clustering in
one area; two groups of suspect samples (Non-conform-1 and Non-conform-
2) cluster in a different region of the plot. This implies that the non-conform
clusters have different volatile profiles from each other, as well as having a differ-
ent volatile profile from the reference samples.
To improve clustering of similar samples, various types of data transforma-
tions and preprocessing can be performed in the exploratory mode prior to creat-
ing classification or regression models. Also, elimination of outliers can be per-
formed to improve modeling results.
The ultimate goal of most multivariate analyses is to develop a model to
predict a property of interest. That property may be categorical or continuous.
Continuous properties are modeled and predicted by regression methodsPrinci-
pal Component Regression (PCR) and Partial Least Squares (PLS), in the case
of Pirouette. After performing exploratory testing, the next step is to decide
FIGURE 4 PCA scores plot showing cluster of normal reference raw materials and two
clusters of suspect raw materials.
whether the goal is to classify unknown samples as to type (i.e., how closely
unknown samples compare to samples of known type) or whether the goal is to
predict a continuous quantitative property (for example, see shelf life prediction
of milk later in this chapter).
To solve a classification problem, the software must first be calibrated for
the particular samples tested. For example, you may be attempting to classify an
unknown ground coffee sample into one of the following categories: Sumatra,
Guatemala, Ethiopia, or French roast. You could use either the KNN (K Nearest
Neighbor) or SIMCA (Soft Independent Modeling of Class Analogy) classifica-
tion techniques for this purpose. First, 1020 known samples of each class type
would be analyzed, and a KNN or SIMCA model would be generated from the
mass intensity lists of these known samples. After running and saving your KNN
or SIMCA model, the next step would be to open the model and analyze unknown
samples. An example of SIMCA Pirouette classification results for a group of
unknown samples appears in Fig. 5. Notice that not only are class assignments
made, but also there is information provided that indicates the quality of the fit
to the model. For details regarding KNN and SIMCA classification techniques,
HCA, PCA, PCR, and PLS, refer to the Pirouette instruction manual or a multi-
variate analysis reference book (10,11).
B. Typical Application Example

As a typical e-nose example, consider the challenge of classifying wines ac-
cording to the vineyard from which they were produced. Can an e-nose instru-
356 Marsili
FIGURE 5 An example of SIMCA results obtained with the Agilent 4440, showing classi-
fication assignments for ground coffee samples.
ment perform as well as human wine tasters? In one case, 68 amateur wine tasters
from Europe, Canada, Latin America, and the United States were given three
glasses of Cabernet Sauvignon; two were from the same vineyard and one was
from a different vineyard. The tasters were asked which wine was different (A,
B, or C) or if they were all the same. Of the 68 people that tested the wine, 23
(34%) gave the correct answer.
When the Agilent 4440 was used to classify the 30 wines according to one
of five vineyards, it accurately classified 100% of the wines (Fig. 6). The Agilent
4440 significantly outperformed amateur wine tasters and won the wine-tasting
challenge.
C. Refinements in Second-Generation Instruments

1. Chemical Ionization Techniques
Positive Chemical Ionization (PCI) can be used in place of Electron Impact (EI)
to improve modeling and class groupings. Consider, for example, 10 different
peppermint oils analyzed using the Agilent 4440 chemical sensor in the EI and
PCI modes.
Figure 7 summarizes the ionization processes normally observed with EI
and PCI using the two most common reagent gasesmethane and ammonia. As
FIGURE 6 PCA scores plot obtained with the Agilent 4440, showing classification of
wine samples according to vineyard (five different vineyards).
FIGURE 7 Mass spectral ionization modes with typical ions resulting from fragmentation
or adduct formation.
358 Marsili
FIGURE 8 Typical fingerprint mass spectra for a peppermint oil obtained using an Agilent
MS chemical sensor in three different modes: (a) electron impact ionization, (b) positive
chemical ionization with methane reagent gas, and (c) positive chemical ionization with
ammonia reagent gas.
shown for typical peppermint oil fingerprint spectra (of all coeluting components)
in Fig. 8ac, electron impact gives the most fragmentation. Because PCI is a
softer ionization technique, much less fragmentation is seen. Methane PCI typi-
cally gives some fragmentation along with an increase in molecular ion adducts
(Fig. 8b). Ammonia PCI usually gives only pseudo molecular ions (Fig. 8c).
PCA analysis (EI mode) of the 10 peppermint oils (five replicates of each)
shows that they fall into eight distinct classes (Fig. 9a). A useful measure of class
separation is the SIMCA interclass distance. As a rule of thumb, interclass distances
of 3 or more indicate good discrimination. Values less than 3 indicate that further
method optimization may be required to obtain reliable predictions. Table 1 shows
some typical interclass distances for the eight peppermint oil categories. The aver-
age of all the interclass distances was 17.7, indicating excellent class separation.
The same peppermint oil samples were then analyzed using methane and
ammonia PCI. Figure 9b shows a PCA scores plot from the NH 3-PCI analysis
FIGURE 9 PCA scores plots of 10 peppermint oils (five replicates of each) showing how
samples fall into eight distinct classes. Samples analyzed with an Agilent MS chemical
sensor operated in two different modes: (a) EI mode and (b) NH 3-PCI mode.
that separates the oils into the same eight classes that were seen using EI (Fig.
9a). As shown in Table 1, the average interclass distance for the NH 3-PCI analy-
ses was 126.6, or more than seven times greater than was obtained using the
Agilent 4440 in the EI mode. NH 3-PCI could produce a much more robust model
with even less chance of misclassification.
TABLE 1 Typical SIMCA Interclass Distances using

Two Different Chemical Sensor Configurations
NH3-PCI
Class pair EI mode mode
12 3.4 36.8
25 8.3 160.1
17 39.3 180.0
24 11.9 270.9
37 25.6 132.8
78 23.2 119.3
Average of all 17.7 126.6

interclass distances
360 Marsili
One useful way to evaluate chemical sensor results and compare EI and CI
is to overlay the normalized spectra for two or more of the classes. It would be
impossible to overlay mass spectra using the conventional bar graph format. Pirou-
ette software converts these plots to line plots, making interpretation of overlaid
plots possible. Figure 10 shows overlaid spectra for 10 different peppermint oils
analyzed in the EI (Fig. 10a) and NH 3-PCI (Fig. 10b) modes. While the entire
m/z range is typically plotted, this plot shows a magnified view in the m/z 148
161 range. These X-residual plots are another way to show that the differences in
peppermint oil responses are much larger with the NH 3-PCI mode than with the
conventional Agilent 4440 chemical sensor in the EI mode. This correlates well
with the increase in SIMCA interclass distances seen when using PCI.
2. Finding Ions That Differentiate Two Samples

Conventional e-noses that employ solid-state sensors can often tell when two
samples are different from each other. In many cases, however, it is important
FIGURE 10 Overlaid fingerprint mass spectra from m/z 148 to m/z 161 for 10 different
peppermint oil samples using an Agilent MS chemical sensor operated in two different
modes: (a) EI mode and (b) NH 3-PCI mode.
to know why they are different. This is particularly true when one sample is
unacceptable for its intended use. A distinct advantage of using MS as a chemical
sensor is that it can be used in tandem with GC/MS to identify the compounds
that differentiate two samples.
Combining the sensor with GC/MS to identify impurities is a three-step
process:
First, X-residual plots from the chemical sensor are used to identify ions
that distinguish the bad sample from the good standard.
Next, samples are analyzed by GC/MS, and these ions are extracted from
the total ion chromatogram (TIC).
Finally, the chromatographic peak(s) that have been pinpointed by the sen-
sor can be identified using conventional library searching. Knowing
which ions are associated with the offending compound(s) makes it much
easier to find and identify them.
3. Automated Calibration Transfer (ACT)

Agilent Technologies and Infometrix have developed a technique called Auto-
mated Calibration Transfer (ACT) that allows users to transfer e-nose applica-
tions from one instrument to another simply by running a few calibration samples.
Running ACT also adjusts for long-term sensor drift and for tune-induced shifts.
In brief, ACT improves the stability and reliability of the MS sensor.
III. ANY GC/MS WILL WORK

A. SPME-MS-MVA
It is noteworthy that the Agilent 4440 is not the only way to perform MS-based
e-nose testing. In fact, any conventional GC/MS can be used with surprisingly
little alteration. Multivariate software other than Pirouette and ion trap mass spec-
trometers (12) can also be used. Most important, static headspace does not have
to be used. Other more sensitive techniques for extracting/concentrating volatiles
can be employedas long as the technique is rapid and, preferably, automatable.
One approach recently reported is referred to as SPME-MS-MVA (13,14).
This technique uses solid-phase microextraction (SPME), mass spectrometry
(MS), and multivariate analysis (MVA) as an e-nose system. A conventional GC/
MS analytical capillary column can be used at an elevated temperature in place
of the short uncoated deactivated retention gap. Coelution of volatile components
occurs, but this is of no concern when using the GC/MS as an e-nose. With this
configuration, one can switch from using the GC/MS as an e-nose to a conven-
tional GC/MS simply by changing the temperature program of the column.
Using a column in place of the uncoated deactivated retention gap provides
two main advantages: (1) No time is wasted converting the instrument to a con-
362 Marsili
ventional GC/MS (as long as the same column is acceptable for both purposes)
and (2) when needed, some temperature programming could be used to separate
analytes. Some chromatographic separation might help to uncover minor differ-
ences in trace components that could otherwise be masked by high concentrations
of other constituents. Chemometric classification might be improved by focusing
on a portion of the chromatogram. Also, with a normal GC capillary column,
it would be possible to operate the system as a conventional GC/MS.
SPME-MS-MVA applications reported to date have used the Varian Saturn
ion trap mass spectrometer and 75-m Carboxen/PDMS as the SPME fiber
(13,14). In one study, for example, SPME-MS-MVA was used to classify various
types of food samples according to the level of oxidized off-flavors they contained
(14). Mass fragmentation data resulting from the unresolved food volatile compo-
nents were subjected to MVA. The mass intensities from m/z 50 to m/z 150
were selected to perform MVA. PCA based on SPME-MS-MVA provided rapid
differentiation of the following types of samples: control soybean oil from oxi-
dized soybean oil that was exposed to fluorescent light for various time periods;
control nondairy coffee creamer from complaint (oxidized) nondairy coffee
creamer samples; fresh boiled beef from boiled beef with various levels of
warmed-over flavor (WOF); and control 2% reduced-fat milk samples from 2%
reduced-fat milk samples abused by light or copper exposure.
SPME is a rapid, solventless extraction/concentration technique that af-
fords significantly lower detection levels for higher molecular weight/higher boil-
ing point compounds than static headspace. Its many advantages over other sam-
ple preparation techniques for flavor, fragrance, and odor analysis have been
pointed out in numerous chapters in this book.
B. Warmed-Over Flavor (WOF )

To illustrate the utility of SPME-MS-MVA for QC food applications, consider
how it can be applied to monitoring the development of WOF in boiled beef.
A beef sample (500 g of chuck roast) was boiled for 60 minutes in a water
bath. The internal temperature of the beef reached 92C. Immediately after boil-
ing, the hot meat was ground in a meat grinder, split into six separate samples,
and analyzed by SPME-MS-MVA. After storage at 4C for 4 days, the samples
were reheated to 50C in a convection oven for 30 minutes. Organoleptic evalua-
tion of the samples showed that their flavor had changed from a typical meaty
beef flavor to an off-flavor characterized as tallowy, green, and metallic. Samples
were again refrigerated, stored for an additional 48 hours, and reanalyzed after
warming to 50C. Samples after 6 days of storage developed even stronger WOF
notes.
Grosch et al. prepared beef samples in a similar way and analyzed samples
using GC-MS equipped with an olfactometry detector (15). Grosch extracted the
meat samples with diethyl ether, and the volatile fractions were distilled off in
vacuo from the nonvolatile material. The volatile fractions obtained from fresh
and from stored boiled meat were subjected to Aroma Extraction Dilution Analy-
sis (AEDA).
Results showed that off-flavor contributions from hexanal, 2-octen-3-one,
(Z)-2-octenal, (Z)-2-nonenal, (E,E)-2,4-nonadienal, and trans-4,5-epoxy-(E)-2-
decenal increased during the storage period, indicating that these compounds
were the most significant contributors to the WOF.
When samples of boiled beef were analyzed fresh (zero days) and after 6
days of storage by SPME-GC-MS, chromatograms revealed many of these same
compounds. Figure 11 shows chromatograms of fresh boiled beef versus boiled
reheated beef after 6 days of refrigeration. A 30 m 0.25 mm FFAP column
was used for the analysis of the meat samples. Converting from a GC/MS e-nose
mode to a conventional GC/MS to perform detailed analysis of individual vol-
atiles was accomplished simply by changing the column temperature program
FIGURE 11 Development of WOF in cooked beef. SPME-GC-MS chromatogram of

boiled beef at zero days and after 6 days of storage and then reheated to 50C. Peak
identities are as follows: (1) pentanal, (2) hexanal, (3) heptanal, (4) 2,4-nonadienal, (5) oc-
tanal, (6) 2,3-octanedione, (7) nonanal, (8) 1-octen-3-ol, (9) 2-heptenal.
364 Marsili
FIGURE 12 PCA scores plot of mass intensity data for fresh boiled beef and boiled beef
refrigerated for 4 days and 6 days and then reheated. Analyses performed by SPME-MS-
MVA.
to one that uses a lower starting temperature. Figure 11 confirms that the key
volatiles responsible for WOF are being extracted by SPME.
Figure 12 shows that SPME-MS-MVA is capable of quickly identifying
groups of samples with similar levels of WOF.
C. Shelf-Life Prediction of Milk

Besides predicting categories for samples, MS e-nose instruments can also be
used for determining a continuous property of samples. Continuous properties
are modeled and predicted by regression methods. Details describing how
SPME-MS-MVA has been used for predicting the shelf life of milk are de-
scribed below (16).
Traditionally, the shelf life of milkthe period between processing and
the time when milk becomes unacceptable to consumers because of taste or
odorhas been determined by bacterial counts and sensory evaluation. The shelf
life of pasteurized processed milk is generally 1418 days, but can be much less
if the milk supply is contaminated with microbes (e.g., psychrotrophic bacteria).
If pasteurized milk with abnormally short shelf life could be detected prior to
leaving the processing plant, there would be less chance that processed milk with
off-flavors would reach consumers tables.
Standard microbiological methods, including the Moseley keeping-quality

test, have proved to be of little value as predictors of shelf life. One problem
with the Moseley keeping-quality test is the extended time period required to
perform testing (5 days). Another problem is the relatively poor correlation be-
tween microbial counts and actual shelf life.
As a result of these problems, dairy chemists and microbiologists are in-
creasingly investigating techniques that measure chemical changes produced by
bacteria rather than measuring total numbers of bacteria. The most popular of
these techniques is electrical impedance. A popular rapid technique for assessing
the number of bacteria present in milk is adenosine triphosphate (ATP) measure-
ments using firefly luciferase and cofactors to produce light. However, these tests
correlate better to total bacterial counts than they do to actual product shelf life
because they do not necessarily measure the direct cause of off-flavor formation
(e.g., malodorous bacterial metabolites) and the end of shelf life.
Recent advances in sample preparation techniques now make analysis of
off-flavor metabolites simple, fast, accurate, and sensitive. With the SPME-MS-
MVA technique, a mass intensity list representing all the volatiles in the milk
sample is the basis for shelf-life predictionnot GC peak area data. SPME-MS-
MVA prediction of processed milk shelf life is based on measurement of volatiles
and semivolatiles present in milk after a pre-incubation period.
1. Sampling
All samples were commercially pasteurized and homogenized reduced-fat milk
(2% milkfat) free of off-flavors at time of manufacture. Samples were packaged
in either pint or half-pint high-density polyethylene (HDPE) contoured bottles
with screw caps. Thirty samples of milk were sampled consecutively from the
production line at a dairy plant the day of processing. This sampling scheme was
conducted on six occasions over a 7-month period.
Samples were immediately taken from the dairy plant and refrigerated in
a walk-in cooler at 7.2 0.5C until the end of shelf life. During refrigerated
storage, two bottles of milk were removed for testing at predetermined inter-
valsthree times weekly in the initial stage of refrigerated storage and then daily
when a decline in flavor quality was observed. One sample from each pair was
subjected to organoleptic evaluation, and one sample was placed in a 19 1C
incubator for 16 hours. After 16 hours, the pre-incubated sample was subjected
to SPME-MS-MVA analysis.
2. Sensory Evaluation
Four judges experienced in tasting dairy products were used for sensory evalua-
tion of milk samples. The method used for sensory scoring was based on a 10-
point scale according to the scoring guide of the American Dairy Science Associ-
366 Marsili
ation. Shelf life was ended when a score of five or lower was recorded by three
of the four judges, and the day before was considered the end of shelf life.
3. SPME Analysis
A Varian Saturn 3 GC/MS was used. The GC was equipped with a split/splitless
model 1078 injector. The injector was operated in the split mode (6 :1 split ratio)
at a temperature of 275C. The SPME fiber used was 75-m Carboxen/PDMS.
For thermal desorption, the SPME fiber remained in the injector for 3 minutes.
Helium was used as the carrier gas. A 30 m 0.25 mm I.D. DB-5 fused-silica
capillary column with a film thickness of 1 m was used, and the flow rate of
the helium carrier gas was 1.0 mL/minute. The following column temperature
programming sequence was used: An initial temperature of 150C was main-
tained for 4 minutes, increased to 180C at a rate of 15C/minute, and held at
180C for an additional 2 minutes. All milk volatile peaks eluted within 7 min-
utes, with many components coeluting.
The objective was to transfer extracted volatiles from the SPME fiber to
the MS in a relatively short time period, rather than waiting approximately 1 hour
for a high-resolution chromatographic run. With this approach, more samples per
hour can be tested. If a sample with unacceptable shelf life is discovered and
more details about specific volatiles are desired, the sample can easily be retested
to improve peak resolution by using a column temperature sequence that starts
at a lower temperature (e.g., 50C) and uses more gradual temperature ramps.
The same analytical column can be used for both approaches, so no time is lost
to column changeover and MS shutdown.
Three milliliters of milk sample, 5 L of internal standard solution (10 g/
mL chlorobenzene), and a micro-stirring bar were placed in a 6 mL-glass GC
vial and capped with PTFE/silicone septa. With the fiber exposed, the sample
vial was placed in a 50C water bath for 20 minutes (fiber exposure started imme-
diately with the sample at 19C), and the sample was stirred at 350 rpm. Multiple
SPME setups initiated at 10-minute intervals significantly increased the number
of samples that could be run per hour.
4. MS Analysis
The Varian Saturn MS detector was used in the electron impact (EI) mode with
a 1-second scan time. The mass range used was m/z 50 to m/z 150. The tempera-
ture of the ion trap manifold was 180C.
A mass intensity list was obtained for each sample by averaging the masses
between 100 s and 500 s. The mass intensities were then normalized by dividing
by the intensity of the major mass peak for the chlorobenzene internal standard
(m/z 112). For PLS calculations, these normalized mass ratios from m/z 50 to
m/z 150 were used as independent variables, and the shelf life determined by
sensory analysis was used as the dependent variable. Internal standard normaliza-
tion of mass intensity data and generation of a normalized mass intensity file in
a .csv file format suitable for conversion to a Pirouette spreadsheet format
were accomplished automatically by a program (Listfile) written in the Varian
Saturn systems application-specific programming language called Procedure
Language. From within the Procedure Language, the user may program specific
actions to accomplish tasks that would otherwise have to be performed man-
ually.
5. MVA Analysis
The software used for MVA was Pirouette from Infometrix, Inc. (Woodinville,
WA). Prediction of shelf life was based on the Partial Least Squares (PLS)
method. Correlation between predicted and actual shelf life of samples was opti-
mum when the following PLS parameters were used: Exclusion of masses 59,
73, 77, and 150; mean centering data preprocessing; Log 10 and SNV data trans-
formation; and 17 model factors. Explanation of these parameters can be found
in the Pirouette manual (10). Masses 73 and 77 were excluded because these are
significant mass peaks that appear in extraneous background compoundse.g.,
hexamethylcyclotrisiloxane, octamethylcyclotetrasiloxane, and decamethyl-
cyclopentasiloxane (from GC septa and degradation of GC column liquid phase)
and fluorotrimethylsilane (a component of the SPME fiber).
SPME-MS-MVA shelf-life prediction models were developed for reduced-
fat milk samples of known shelf life. Mass intensity lists were determined for
84 samples of reduced-fat milk. Sixty-four of these samples were used to develop
a PLS calibration model, and 20 samples (a Model Validation Subset) were
randomly selected from the set of 84 total samples to evaluate how well the PLS
model for reduced-fat milk could predict shelf life.
6. Results and Conclusions

Poor correlation of microbiological plate counts to shelf life is likely due to two
factors. First, spoilage is not always related to the number of organisms present. The
type of bacteria present rather than the actual numbers determines the types and the
extent of off-flavor development and consequently the end of shelf life. Second, poor
flavor and shelf life can also be attributed to the presence of microbial enzymes and
metabolic products from organisms present before pasteurization. In some cases,
pasteurization can kill microorganisms but not inactivate microbial enzymes.
Furthermore, volatiles produced as bacterial metabolites or from active mi-
crobial enzymes are not the only causes of off-flavors in milk and shortened
shelf life. For example, chemical contaminants can generate significant off-flavor
formation. Contamination of milk with relatively low levels of copper, for exam-
ple, can result in formation of oxidation off-flavors during shelf life. Another
nonmicrobiological cause of off-flavors that sometimes occurs in milk samples
that first come off the processing line is contamination by sanitizer from pro-
368 Marsili
cessing lines that havent been properly flushed. It is noteworthy that SPME-
MS-MVA testing is able to detect and identify these nonmicrobiological sources
of off-flavors and decreased shelf life.
Table 2 compares actual shelf life (determined by sensory evaluation) to
predicted shelf life for the 20 PLS Model Validation Subset samples for reduced-
fat milk. On average, the SPME-MS-MVA PLS model for reduced-fat milk pre-
dicted the shelf life with an accuracy of 0.62 days, with a correlation coefficient
of 0.9801 and a range of 0.7 to 2.8 days.
TABLE 2 Actual a and Predicted b Shelf Life of

Homogenized, Pasteurized Reduced-Fat Milk
Shelf life in days
Actual Predicted Error c

15 15.3 0.3
18 18.2 0.2
14 14.7 0.7
8 7.4 0.6
5 5.4 0.4
1 1.8 0.8
14 13.6 0.4
11 10.3 0.7
10 10.2 0.2
7 6.4 0.6
4 5.0 1.0
3 4.2 1.2
2 2.3 0.3
0 0.7 0.7
10 12.8 2.8
7 7.3 0.3
3 3.5 0.5
2 2.6 0.6
5 4.9 0.1
0 0.0 0.0
Average error d: 0.62
Error range: 0.7 to 2.8
R 2: 0.9801
a
Determined by sensory panel.
b
Predicted from SPME-MS-MVA data using PLS pre-
diction models.
c
Error predicted actual.
d
Average error (| error |)/n, where n 20.
FIGURE 13 Plot of predicted shelf life (based on PLS modeling) versus actual shelf life
(based on sensory testing) for 64 samples used to prepare a PLS model for reduced-fat
milk. Analyses performed by SPME-MS-MVA.
Figure 13 shows a plot of predicted shelf life versus actual shelf life (based
on sensory testing) for the 64 samples used to prepare the PLS model for reduced-
fat milk. Table 3 shows PLS statistics for the PLS shelf-life prediction model
and the PLS predictions of Model Validation Subset samples.
Preliminary results using SPME-MS-MVA as an electronic-nose system
appear to give more accurate predictions of milk shelf life than most methods
currently used to estimate milk shelf life. (See Table 4.) SPME-MS-MVA is also
faster and easier to implement than other shelf-life prediction methods. Further-
more, SPME-MS-MVA has been shown to be useful for identifying samples
with nonmicrobiological induced off-flavors and for determining the cause of
off-flavors even when nonmicrobiological agents are involved.
Over a 7-month period, SPME-MS-MVA has been shown to be an accurate
technique for predicting the shelf life of reduced-fat milk. Despite the fact that
during the testing period significant changes occurred with the mass spectrometer
(replacement of the turbomolecular pump and replacement of the electron multi-
plier) and the fact that several different Carboxen/PDMS fibers were used, inter-
nal standard normalization with chlorobenzene allowed accurate prediction over
the 7-month period. Long-term stability, a problem with many e-nose instruments
based on solid-state sensors, does not appear to be a significant problem with
MS-based e-nose instruments.
Using Carboxen/PDMS SPME fibers to extract volatiles offers impressive
370 Marsili
TABLE 3 PLS Error Analysis for Calibration Model

and Model Validation Subset for Reduced-Fat Milk
(shelf life in days as dependent variable)
Calibration Model validation
model subset
PRESS a 53.4204 25.5115
SEC b (days) 1.0441
SEP c (days) 1.1294
R2 0.9882 0.9801
Factors 15 7
Slope 0.9766 0.9492
Intercept (days) 0.2014 0.0750
n 64 20
a
PRESS Prediction Residual Error Sum of Squares.
b
SEC Standard Error of Calibration (PRESS/n) 1/2.
c
SEP Standard Error of Prediction [PRESS/(n k)] 1/2,
where k number of factors.
advantages over static headspace (SH) and dynamic headspace (DH) sampling
techniques. It does not require expensive ancillary instrumentation and is far more
efficient than either SH or DH at extracting volatile fatty acids (VFAs) from
milk. VFAs, important contributors to malodors and off-flavors in milk, are gen-
erated as metabolites by the growth of lipolytic psychrotrophic bacteria. Malodor-
ous VFAs are too polar to detect at low levels using SH and DH as sample
preparation/extraction tools.
TABLE 4 Correlation Coefficients for Various Shelf-Life

Prediction Tests
Correlation
coefficient
Shelf-life prediction test (predicted vs. actual)
Moseley Keeping-Quality Test 0.70.77

Catalase 0.77
ATP 0.880.92
DEFT a 0.72
VTSL b 0.89
SPME-MS-MVA 0.98
a
Direct Epifluorescent Filter Technique.
b
Virginia Tech Shelf-Life Program.
Source: Dr. Russel Bishop, University of Wisconsin, Madison, Wisconsin.
SPME-MS-MVA has strong potential applications in the dairy industry for

shelf-life prediction. Testing over a longer period of time and sampling from
different production facilities should be conducted to confirm the accuracy of
this new test as a predictor of shelf life. With a SPME autoinjector and minor
test modifications, it would be possible to analyze one sample every five to seven
minutes. The only labor required by the QC technician would be to pipette 3 mL
of milk into a GC vial.
A simple, rapid, and sensitive sample preparation technique such as SPME
and applying GC/MS in a nontraditional way are the basis of this new shelf-life
prediction test for milk. This approach could be extended to many other types
of important quality control applications in the food industry as well as in other
industries.
IV. FUTURE DIRECTIONS
What would constitute the ideal MS-based e-nose system? It would likely: (a) in-
corporate a more sensitive technique than static headspace to deliver volatiles to
the sensor array; (b) be less expensive than most e-nose instruments currently
on the market; (c) be easy to use; and (d) provide results in less than 10 minutes
per sample. An optimum instrument configuration would also allow the same
instrument to be used in a rapid e-nose mode but would also permit investigation
of sample anomalies using conventional GC/MS methods. Such an instrument
could be used as a rapid screening tool and also as a research tool for uncovering
further chemical information about suspect samples. Switching from one mode
to the other should not require any hardware modification or even instrument
shutdown to change columns.
One promising approach that meets these criteria is a refinement in the
SPME-MS-MVA shelf-life prediction technique described above. Details are
shown in Fig. 14 for a dairy QC application. This strategy uses SPME-MS-MVA
as an initial screening tool to detect processed milk samples with unusually short
shelf life. The approach extends the sophistication of SPME-MS-MVA by incor-
porating an SPME autosampler in addition to a fast-GC column. Varians pat-
ented approach to fast-GC is called Fast-MS and is capable of reducing chro-
matographic run times fivefold or greater without significant loss in peak
resolution. Using Fast-MS, analyses could be performed with the speed of the
Agilent 4440 chemical sensor equipped with a short, uncoated deactivated reten-
tion gap but still provide peak resolution nearly equivalent to GC/MS with tradi-
tional capillary columns.
Figure 14 illustrates a possible strategy for implementing SPME-MS-MVA
in a dairy QC lab. A processed milk sample would first be analyzed in the dairy
processing plant laboratory by SPME-MS-MVA to estimate shelf life as de-
372 Marsili
FIGURE 14 A strategy for using SPME-MS-MVA as a dairy QC screening tool. Rapid

screening is performed using a GC/MS as an e-nose to estimate shelf life of fresh pro-
cessed milk; subsequent checking of suspect samples (i.e., those with unusually short shelf
life) can be conducted with additional manipulation of the TIC GC/MS file by trained
chemists (e.g., at a corporate research chemistry lab) without the need for retesting sam-
ples. *MVA based on mass intensity data.
scribed above, with the only modifications being the use of an autosampler (the
Combi PAL from CTC Analytics, Switzerland) and Rapid-MS. A low-cost Var-
ian Saturn 2100 is specified in Fig. 14. This instrument is a smaller benchtop
GC/MS that couples the Saturn ion trap mass spectrometer with a new GC that is
small, low cost, and yet provides all the performance, MS options, and analytical
capabilities of most standard GC/MS systems.
If the initial screening indicates a potential shelf life problem (for example,
if SPME-MS-MVA predicted a shelf life of 10 days or less instead of the typical
1517 days), then the chromatographic file (total ion chromatogram) could be
subjected to further scrutiny in order to uncover the cause of the off-flavor. The
.ms file, which contains all the information necessary to perform conventional
MS identification of chromatographic peaks, could be e-mailed to a corporate
research lab. The corporate lab could then perform more sophisticated analysis
of the data. It could, for example, do further multivariate analysis investigations
(e.g., PCA and HCA) to classify the suspect milk sample as to the type of abuse
the sample has experienced (bacterial spoilage, overheating during pasteurization,
contamination by sanitizer, oxidation caused by contact with pro-oxidant metals,
etc.). Also, the corporate lab could perform traditional MS peak identification
(MS library matching) to identify the key chemicals responsible for the off-flavor
and deduce the mechanism of off-flavor formation. It is important to note that
further examination and interpretation of the high-resolution chromatogram can
be accomplished without the need to reanalyze the sample. Detailed investiga-
tions are performed on the total ion chromatograms generated by the QC labs
GC/MS original analysis.
V. CONCLUSION
MS-based e-nose instrumentation will likely grow in popularity as a routine

screening tool for QC monitoring of flavors, fragrances, and odors in raw materi-
als, packaging, and finished products; the advantages of MS-based e-nose instru-
ments are many:
Less drift (proven bench-top MS technology)

Fast (25 minutes per sample)
No problems with water
No problems with alcohols, sulfur-containing chemicals, and polar com-
pounds
Scan range determines the number of sensors
Linear to 10 4
No poisoning of sensors
Correlation with GC/MS
Capable of identifying the ions that differentiate two samples
GC/MS-based approach offers:
more sample introduction techniques
chemical ionization
conventional GC/MS testing when needed
One of the most important benefits of using MS as a sensor array in an

e-nose instrument is that it can be used to determine not only if a test sample is
different from a standard sample but also why it is different. The ability to use
MS e-nose instruments as both a rapid screening productivity tool and a research
tool to provide more details about specific chemical components in samples is
an appealing combination.
374 Marsili
ACKNOWLEDGMENTS
The author wishes to express appreciation to Dr. Phil Wylie of Agilent Technolo-
gies for providing the figures showing application examples of the Agilent 4440
chemical sensor.
REFERENCES
1. H. Kim-Kang, Volatiles in packaging materials, Crit. Rev. Food Sci. Nutri. 29:255
(1990).
2. N. Goldenberg and H. R. Matheson, Off-flavors in foods, a summary of experiences:
19481974, Chem. Industry 5:551 (1975).
3. R. T. Marsili, Measuring light-induced chemical changes in soybean oil by capillary
headspace GC, J. Chromatogr. Sci. 22:61 (1984).
4. R. T. Marsili, G. J. Kilmer, and N. Miller, Quantitative analysis of orange oil compo-
nents in orange juice by a simple solvent extraction-GC procedure, LC GC 7:778
(1989).
5. D. Hodgins, The electronic nose: sensor array-based instruments that emulate the
human nose, Techniques for Analyzing Food Aromas (R. Marsili, ed.), Marcel Dek-
ker, New York, 1997, p. 331.
6. A. M. Pisanelli, A. A. Qutob, P. Travers, S. Szysko, and K. C. Persaud, Applications
of multi array sensors to food industries, Life Chemistry Reports 11:303 (1994).
7. K. C. Persaud, Odor detection using sensor array, Anal. Proc. 28:339 (1991).
8. K. C. Persaud and P. Pelosi, Sensory arrays using conducting polymers for an artifi-
cial nose, presented at a NATO workshop on electronic noses, Reykjavik, Iceland,
1991.
9. J. W. Gardner and P. N. Bartlett, Electronic Noses: Principles and Applications,
Oxford University Press, New York, 1999.
10. Pirouette: Multivariate Data Analysis for Windows 95/98 and NT, Infometrix, Inc.,
Woodinville, WA.
11. K. R. Beebe, R. J. Pall, and M. B. Seaholtz, Chemometrics: A Practical Guide, John
Wiley & Sons, Inc., New York, 1998.
12. C. K. Huston, Ion trap mass spectrometry for food aroma analysis, Techniques for
Analyzing Food Aromas (R. Marsili, ed.), Marcel Dekker, 1997, p. 209.
13. R. T. Marsili, SPME-MS-MVA as an electronic nose for the study of off-flavors in
milk, J. Agric. Food Chem. 47:648 (1999).
14. R. T. Marsili, SPME-MS-MVA as a rapid technique for assessing oxidation off-
flavors in foods, presented at the 216th ACS National Meeting, Boston, August 23
27, 1998.
15. W. Grosch, U. C. Konopka, H. Guth, In Lipid Oxidation in Food (A. J. St. Angelo,
ed.), ACS Symposium Series No. 500, American Chemical Society, 1989, p. 266.
16. R. T. Marsili, Shelf-life prediction of processed milk by solid-phase microextraction,
mass spectrometry, and multivariate analysis, J. Agric. Food Chem. 48:3470 (2000).
14
Character Impact Compounds:
Flavors and Off-Flavors in Foods
Robert J. McGorrin
Oregon State University, Corvallis, Oregon
I. INTRODUCTION
The aroma substances that comprise flavors are found in nature as complex mix-
tures of volatile compounds. A vast majority of volatile chemicals that have been
isolated from natural flavor extracts do not provide aroma contributions that are
reminiscent of the flavor substance. For instance, n-hexanal is a component of
natural apple flavor (1); however, when smelled in isolation, its odor is reminis-
cent of green, painty, rancid oil. Similarly, ethyl butyrate has a nondescript
fruity aroma; although it is found in strawberries, raspberries, and pears, it
does not uniquely describe the aroma quality of any of these individual fruits. It
has long been the goal of flavor chemists to elucidate the identity of pure aroma
chemicals that have the distinct character impact of the natural fruit, vegetable,
meat, cheese, or spice that they were derived from. Often, these are referred to
as character impact compounds (2).
The character impact compound for a particular flavor or aroma is a unique
chemical substance that provides the principal sensory identity. Often, character im-
pact is elicited by a synergistic blend of several aroma chemicals. When tasted or
smelled, the character impact chemical, or group of chemicals, contributes a recog-
nizable sensory impression even at low concentration levels as typically found in
natural flavors (for example, vanillin in vanilla extract, and diacetyl in butter) (3,4).
In some instances, flavor concentration and food context are very important. For
example, at high concentrations, 4-mercapto-4-methyl-2-pentanone (cat ketone)
has an off-odor associated with cat urine, but in the context of a Cabernet Sauvignon
wine, it provides the typical flavor impression of the Sauvignon grape (5).
375
376 McGorrin
For many foods, character impact compounds are unknown or have not
been reported to date. Examples of these include Cheddar cheese, milk chocolate,
and sweet potatoes. For these foods, the characterizing aroma appears to be com-
posed of a relatively complex mixture of flavor compounds, rather than one or
two aroma chemicals.
The intent of this chapter is to summarize what is generally known about
the chemical identities of characterizing aroma chemicals in fruits, vegetables,
nuts, herbs and spices, and savory and dairy flavors. A short compendium of
characterizing off-flavors and taints that have been reported in foods is also dis-
cussed.
II. CHARACTER IMPACT FLAVORS IN FOODS
More than 6,000 compounds have been identified in the volatile fraction of foods
(6). The total concentration of these naturally occurring components varies from
a few parts-per-million (ppm) to approximately 100 ppm, with the concentration
of individual compounds ranging from parts-per-billion to parts-per-trillion. A
majority of these volatile compounds do not provide significant impact to flavor.
For example, more than 700 compounds have been identified in the flavor of
coffee, but in general only a small proportion of these substances make a signifi-
cant contribution to the flavor profile (2).
The ultimate goal of flavor research in the food industry is to identify
and classify unique aroma chemicals that contribute to the characteristic
odor and flavor of foods. Having this knowledge enables flavor duplication
through nature-identical or biosynthetic pathways and can facilitate better quality
control of raw materials by screening of the appropriate analytical target com-
pounds.
In recent studies, potent aroma compounds have been identified using vari-
ous gas chromatography-olfactometry (GCO) techniques, such as Charm Analy-
sis and aroma extract dilution analysis (AEDA) (7,8). The flavor compounds that
are identified by these methods are significant contributors to the sensory profile.
In some cases, these sensory-directed analytical techniques have enabled the dis-
covery of new character impact compounds. However, in other instances, key
aroma chemicals have been identified that, while potent and significant to flavor,
do not impart character impact. For example, in dairy products, chocolate, and
kiwifruit, these flavor types appear to be produced by a complex blend of non-
characterizing key aroma compounds.
When character impact compounds are known, flavor chemists are able to
use these materials as basic keys to formulate enhanced versions of existing
flavors. As analytical techniques improve in sensitivity, flavor researchers con-
tinue their quest to discover new character impact flavors that will enable them
to develop the next generation of improved flavor systems.
Character Impact Compounds 377
A. Herb, Spice, and Seasoning Flavors

The original identifications of character aroma compounds were from isolates of
spice oils and herbs. Many of these early discoveries paralleled developments in
synthetic organic chemistry (9). The first identifications and syntheses of char-
acter flavor molecules include benzaldehyde (cherry), vanillin (vanilla), methyl-
salicylate (wintergreen), and cinnamaldehyde (cinnamon). A listing of character
impact compounds found in herb and spice flavors is presented in Table 1.
TABLE 1 Character Impact Flavor Compounds in Herbs, Spices, and Seasonings

Character impact compound(s) CAS registry no. Occurrence Reference
Anethole [4180-23-8] Anise 4
Methyl chavicol (estragole) [140-67-0] Basil 10
d-Carvone [2244-16-8] Caraway 11
trans-Cinnamaldehyde [104-55-2] Cinnamon 4
Eugenol [97-53-0] Clove 4
Eugenyl acetate [93-28-7] Clove 10
d-Linalool [78-70-6] Coriander 4
trans-2-Dodecenal [20407-84-5] Coriander 15
Cuminaldehyde [122-03-2] Cumin 10
p-1,3-Menthadien-7-al [1197-15-5] Cumin 10
(S)--Phellandrene [99-83-2] Fresh dill 10
3,9-Epoxy-p-menth-1-ene [74410-10-9] Fresh dill 17
1,8-Cineole (eucalyptol) [470-82-6] Eucalyptus 11
Sotolone [28664-35-9] Fenugreek 12
Diallyl disulfide [2179-57-9] Garlic 18
Diallylthiosulfinate (allicin) [539-86-6] Garlic 18
1-Penten-3-one [1629-58-9] Horseradish 4
4-Pentenyl isothiocyanate [18060-79-2] Horseradish 18
Allyl isothiocyanate [57-06-7] Mustard 4
Propyl propanethiosulfonate [1113-13-9] Onion, raw 18
3-Mercapto-2-methylpentan-1-ol [227456-30-6] Onion, raw 21
Allyl propyl disulfide [2179-59-1] Onion, cooked 18
2-(Propyldithio)-3,4-Me2thiophene [126876-33-3] Onion, fried 18
Carvacrol [499-75-2] Oregano 10
-t-, c-Bergamotene [6895-56-3] Pepper, black 16
l-Menthol [89-78-1] Peppermint 4
t-4-(MeS)-3-butenyl isothiocyanate [13028-50-7] Radish 18
Verbenone [80-57-9] Rosemary 11
l-Carvone [6485-40-10] Spearmint 11
Safranal [116-26-7] Saffron 14
ar-Turmerone [532-65-0] Turmeric 13
Thymol [89-83-8] Thyme 11
Methyl salicylate [119-36-8] Wintergreen 4
378 McGorrin
A major contrbutor to the flavor of basil is methyl chavicol, which provides

tealike, green, hay, and minty notes (10). In Italian-spice dishes, basil is comple-
mented by oregano, of which carvacrol is a character impact aroma. Thyme
is the dried leaves of Thymus vulgaris, a perennial of the mint family, for which
thymol provides a warm, pungent, sweetly herbal note. In a classic example of
the effect of chiral isomers on flavor character, (S)-()-carvone imparts caraway
flavor, whereas (R)-(-)-carvone provides spearmint flavor (11).
Toasted and ground fenugreek seed is an essential ingredient of curry
powders. Sotolone (3-hydroxy-4,5-dimethyl-2(5H )furanone) was recently estab-
lished as a character impact flavor component in fenugreek on the basis of of
its seasoning-like flavor note (12). Its aroma characteristic changes from
caramel-like at low concentration levels to currylike at high concentrations. The
sensory impact of sotolone is attributed to its extremely low detection threshold
(0.3 mg/l water), and the fact that its concentration in fenugreek seeds is typically
3000-times higher than its threshold.
Turmeric is also primarily used as a spice component in curry dishes and
as a coloring agent in dried and frozen foods. The character impact compound
for turmeric is reported as ar-turmerone (13). Saffron, the dried red stigmas of
Crocus sativus L. flowers, is utilized to impart both color and flavor, which is
described as sweet, spicy, floral, with a fatty herbaceous undertone. Safranal
(2,6,6-trimethyl-1,3-cyclohexadiene-1-carboxaldehyde) has been generally con-
sidered to be the character impact compound of saffron; however, a recent in-
vestigation has also identified two other potent compounds, 4,4,6-trimethyl-2,5-
cyclohexadien-1-one and an unknown, possessing saffron, stale, dried-hay
aroma attributes (14). Representative structures for spice impact compounds are
shown in Fig. 1.
A key component in both chili powder and curry powder, cumin is the dried
seed of the herb Cuminum cyminum, a member of the parsley family. Cuminaldehyde
is the principal contributor to the spices aroma and flavor, which imparts a strong
musty/earthy character, with green grassy notes contributed by p-1,3- and 1,4-men-
thadienals. trans-2-Dodecenal, possessing a persistent fatty-citrus-herbaceous odor,
is a character impact component of coriander, along with d-linalool (15).
Thus far, very little has been reported concerning the importance of sesqui-
terpenes in natural flavors. Although a high proportion of monoterpenes are pres-
ent in black pepper oil with little consensus as to their relative importance (10),
sesquiterpene isomers -trans- and -cis-bergamotene provide a distinct odor of
ground black pepper (16a), while ()-sabinine and linalool play a supporting
role (16b). It is well known that ()--phellandrene, the main constituent of dill
(Anethum graveolus L.), greatly contributes to the sensory impression of dill herb
(10). However, fresh dill character impact appears to be a synergistic relationship
contributed primarily by ()--phellandrene with a modifying effect from dill
ether, 3,9-epoxy-p-menth-1-ene (17).
FIGURE 1 Representative herb and spice character impact flavor compounds.

380 McGorrin
FIGURE 1 Continued
The Allium family includes garlic, onion, leek, and chive. All are composed
of sulfur-containing character impact compounds. The aroma impact constituents
of garlic are diallyl disulfide and the corresponding thiosulfinate derivative (al-
licin), which are enzymatically released from a sulfoxide flavor precursor (alliin)
during the crushing of garlic cloves (18).
The flavor chemistry of sulfur compounds in onion is quite complex
(19,20). Early reports of polysulfides and thiosulfinates were later demonstrated
to be thermal artifacts from gas chromatographic analysis (20). Character impact
sulfur compounds have been proposed for fresh, boiled, and fried onion. In raw,
fresh onion, propyl propanethiosulfinate, propenyl propanethiosulfinate thiopro-
panal S-oxide, and propyl methanethiosulfinate are impact contributors (18,19).
A number of compounds contribute to the aroma character of cooked onion, of
which dipropyl disulfide and allyl propyl disulfide provide key impact (18). Fried
onion aroma is formed by heating the latter compound, and is characterized by
2-(propyldithio)-3,4-dimethylthiophene, which has an odor threshold of 1050

ng/l in water. A recent study identified a new potent aroma compound from raw
onions, 3-mercapto-2-methylpentan-1-ol (21). The flavor impact of this thiol is
strongly dependent on concentration; at 0.5 ppb, it provides a pleasant brothlike,
sweaty, onion, and leeklike flavor; at high levels, it provides a strong, unpleasant
onion-like quality.
Isothiocyanates are character-impact constituents that provide pungency
and typical flavor to mustard (allyl isothiocyanate), radish [trans-4-(methylthio)-
3-butenyl and trans-4-(methylthio)butyl isothiocyanate] and horseradish (4-pen-
tenyl and 2-phenylethyl isothiocyanate) (18).
B. Fruit Flavors
The aroma constituents of essential oils from fruits such as lime, lemon and
orange were among the first character impact compounds identified by flavor
chemists. Fruit flavors are a subtle blend of characterizing volatile compounds,
supported by fruit sugars, organic acids, and non-characterizing volatile esters.
Fruit aromatics tend to be present in concentrations of greater abundance (30
ppm) than other foods, which facilitated early analytical studies. The volatile
composition of fruits is extremely complex, and non-characterizing flavor esters
are common across species. A compilation of character impact compounds found
in fruit flavors is summarized in Table 2.
The combination of ethyl 2-methyl butyrate, -damascenone, and hexanal
is important for the characteristic flavor note of the Delicious apple (22,23a). The
blend of character impact flavors combines apple ester and green apple
notes, which fluctuate with apple ripeness and seasonality. -Damascenone is an
unusually potent aroma compound with a threshold of 2 pg/g in water, and it
also occurs in natural grape and tomato flavors (23a).
Two important character-impact compounds of strawberry flavor are
the furanones 2,5-dimethyl-4-hydroxy-2H-furan-3-one (Furaneol) and 2,5-
dimethyl-4-methoxy-2H-furan-3-one (mesifuran) (24). However, at various con-
centrations, Furaneol can simulate other flavors, e.g., pineapple (11) or Musca-
dine grape (25) at low levels, and caramel at high levels. Mesifuran exhibits a
sherry-like aroma and is a contributor to sherry and French white wine aroma.
Other important character-impact compounds of strawberry flavor are methyl cin-
namate and ethyl 3-methyl-3-phenylglycidate, a synthetic aroma chemical (4,26).
Representative chemical structures for fruit flavor impact compounds are shown
in Fig. 2.
The character impact component for Concord (Labruska) grape has been
long known as methyl anthranilate. More recently, ethyl 3-mercaptopropionate
was identified in Concord grape, and in the low-ppm range it possesses a pleasant
fruity fresh Concord grape aroma (27). 2-Aminoacetophenone and mesifuran are
382 McGorrin
TABLE 2 Character Impact Flavor Compounds in Fruits

Ethyl-2-methyl butyrate [7452-79-1] Apple 4
-Damascenone [23696-85-7] Apple 4
iso-Amyl acetate [123-92-2] Banana 4
4-Methoxy-2-methyl-2-butanethiol [94087-83-9] Blackcurrant 18
iso-Butyl 2-butenoate [589-66-2] Blueberry 4
Benzaldehyde [100-52-7] Cherry 4
Tolyl aldehyde [1334-78-7] Cherry 4
Methyl anthranilate [134-20-3] Grape, Concord 18
Mesifuran [4077-47-8] Grape, Concord 18
Ethyl heptanoate [106-30-9] Grape, wine 4
4-Mercapto-4-Me-2-pentanone [19872-52-7] Grape, Sauvignon 5
Nootkatone [4674-50-4] Grapefruit 4
1-p-Menthene-8-thiol [71159-90-5] Grapefruit 28
Citral (neral geranial) [5392-40-5] Lemon 4
-Terpineol [98-55-5] Lime 4
Citral (neral geranial) [5392-40-5] Lime 4
2,6-Dimethyl-5-heptenal [106-72-9] Melon 11
Z-6-Nonenal [2277-19-2] Melon 22
-Sinensal [8028-48-6] Orange 4
Octanal [124-13-0] Orange 4
Decanal [112-31-2] Orange 4
Methyl N-methylanthranilate [85-91-6] Orange, Mandarin 29
Thymol [89-83-8] Orange, Mandarin 29
3-Methylthio-1-hexanol [5155-66-9] Passion fruit 18
2-Methyl-4-propyl-1,3-oxathiane [67715-80-4] Passion fruit 18
-Undecalactone [104-67-6] Peach 4
6-Pentyl-2H-pyran-2-one [27593-23-3] Peach 4
Ethyl trans-2,cis-4-decadienoate [3025-30-7] Pear, Bartlett 4
Allyl caproate [123-68-2] Pineapple 11
Ethyl 3-(methylthio)propionate [13327-56-5] Pineapple 31
Allyl 3-cyclohexylpropionate [2705-87-5] Pineapple 11
4-( p-Hydroxyphenyl)-2-butanone [5471-51-2] Raspberry 4
trans--ionone [127-41-3] Raspberry 4
Ethyl 3-methyl-3-phenylglycidate [77-83-8] Strawberry 4
Furaneol [3658-77-3] Strawberry 24
Mesifuran [4077-47-8] Strawberry 24
(Z,Z )-3,6-Nonadienol [53046-97-2] Watermelon 22
FIGURE 2 Representative fruit character impact flavor compounds.
also significant contributors to the foxy and candy-like notes, respectively,

in Concord grape. This aroma character is in significant contrast to that of Vitis
vinifera varieties used for production of table wines, in which ethyl heptanoate
has an odor reminiscent of cognac (4). 4-Mercapto-4-methyl-2-pentanone pro-
vides Sauvignon grape character, as discussed earlier (5).
384 McGorrin
Two character impact compounds have been proposed for grapefruit

flavor, the first being nootkatone, a sesquiterpene. The fresh juicy note of
grapefruit juice is attributable to 1-p-menthene-8-thiol. This compound has a
detection threshold of 101 ppt, among the lowest values reported for aroma
chemicals (28). The ()-R-isomer was found to have a lower aroma threshold
in water than the racemic mixture, and it imparts a pleasant, fresh grapefruit juice
character, as opposed to the extremely obnoxious sulfur note contributed by the
()-S-epimer.
Among other citrus flavors, the basic flavor-impact compound of lemon is
citral, a mixture of neral and geranial isomers, which together compose the aroma
impression. The flavor character of lime results from a combination of -terpineol
and citral, even though limonene is the most abundant, but sensorily immaterial,
volatile in lime and other citrus oils. In contrast to other citrus flavors, a single
flavor impact compound is unknown for orange, and current belief is that orange
flavor is the result of a complex combination of terpene, aldehyde, ester, and
alcohol volatiles in specific proportions (29). The most odor-active compounds
in fresh orange include the sesquiterpene aldehyde -sinensal, octanal, and dec-
anal (4). For Mandarin and tangerine flavors, methyl-N-methyl anthranilate and
thymol are associated with character impact, with additional contributions from
-pinene and -terpinene (29).
Lactones have characteristic aromas that contribute to peach, coconut, and
dairy flavors and occur in a wide variety of foods. The -lactones, specifically
-undecalactone and lesser for -decalactone, possess intense peach-like odors
(11). A doubly unsaturated -decalactone, 6-pentyl-2H-pyran-2-one, also has an
intense peach character (4). As a point of distinction, the C10-C12 -lactones, par-
ticularly the creamy-coconut note of -decalactone, are flavor constituents of
coconut as well as cheese and dairy products (11).
In addition to the character impact compound of raspberry, 4-(4-hydroxy-
phenyl)-butan-2-one (raspberry ketone), alpha- and beta-ionone, geraniol, and
linalool were concluded to be of importance in raspberry aroma (30). Its odor
threshold was measured at 110 g/kg. The ionones have chemical structure
similarities and potencies comparable to -damascenone (11).
The tropical category is one of the most important areas for new discov-
eries of key impact flavor compounds. Analyses of passion fruit and durian fla-
vors have produced identifications of many potent sulfur aroma compounds (18).
Among these is tropathiane, 2-methyl-4-propyl-1,3-oxathiane, which has an odor
threshold of 3 ppb (15). For pineapple, 2-propenyl hexanoate (allyl caproate)
exhibits a typical pineapple character (11); however, Furaneol, ethyl 3-methyl-
thiopropionate, and ethyl-2-methylbutyrate are important supporting character
impact compounds (31). The latter ester contributes the background apple
note to pineapple flavor. Another character impact compound, allyl 3-cyclohexyl-
propionate, has not been discovered in nature, but it provides a sweet-fruity pine-
apple flavor note (11).
Blackcurrant flavor is very popular in Europe, and is associated with numer-
ous health-related functional foods and with alcoholic drinks (cassis liqueur). The
key aroma component in blackcurrant is 2-methoxy-4-methyl-4-butanethiol (18).
Characterizing flavors for melons include (Z )-6-nonenal, which contributes a typ-
ical melon aroma impression, and (Z,Z )-3,6-nonadienol for watermelon rind
aroma impact (22). 2,6-Dimethyl-5-hepten-1-al (Melonal) has not been identi-
fied in melon, but it provides a melon-like note in compounded flavors (11). The
flavor of muskmelons is more complex, with methyl 2-methylbutanoate, Z-3-
hexenal, E-2-hexenal, and ethyl 2-methylpropanoate identified as the primary,
non-characterizing odorants (32).
While an important constituent to both varieties, the flavor of sweet cherries
(Prunus aviium) is less dominated by the character impact compound, benzalde-
hyde, than is the flavor profile of sour cherries (P. cerasus) (33).
C. Vegetable Flavors
Recent aroma research has been devoted to the identification of key flavor com-
pounds in vegetables and is the subject of several contemporary reviews
(31,34,35). Cucumbers, sweet corn, and tomatoes are botanically classified as
fruits; however, for flavor considerations they are regarded as vegetables, because
they are typically consumed with the savory portion of the meal. Overall, the
knowledge base of character impact compounds for vegetables is much smaller
than other flavor categories and warrants further investigation.
Identifying flavor impact compounds in vegetables depends considerably
on how they are prepared (cutting, blending) and the form in which they are
consumed (raw vs. cooked). For example, the character impact of fresh tomato
is delineated by 2-iso-butylthiazole and (Z )-3-hexenal, with modifying effects
from -ionone and -damascenone (31). Alternatively, thermally processed to-
mato paste has dimethyl sulfide as a major flavor contributor (31,34). Dimethyl
sulfide is also a flavor impact compound for both canned cream corn and fresh
corn; 2-acetyl-1-pyrroline provides a corn chip character. Other sulfur com-
pounds such as hydrogen sulfide, methanethiol, and ethanethiol may contribute
to the aroma of sweet corn due to their low odor thresholds (35). A summary of
character impact compounds for vegetable flavors is outlined in Table 3.
The character-impact compound of green bell pepper, 2-isobutyl-3-
methoxypyrazine, was the first example of a high-impact aroma compound be-
cause of its exceptionally low odor threshold of 2 parts-per-trillion (ppt) (15). It
is also responsible for the vegetative aroma of Cabernet Sauvignon wine
(23b). A similarly low odor threshold compound, geosmin, is the character impact
386 McGorrin
TABLE 3 Character Impact Flavor Compounds in Vegetables

Dimethyl sulfide [75-18-3] Asparagus 4
1,2-Dithiacyclopentene [288-26-6] Asparagus, heated 34
Geosmin [19700-21-1] Red beet 34
4-Methylthiobutyl isothiocyanate [4430-36-8] Broccoli 35
Dimethyl sulfide [75-18-3] Cabbage 34
Methyl methanethiosulfinate [13882-12-7] Sauerkraut 38
2-sec-Butyl-3-methoxypyrazine [24168-70-5] Carrot (raw) 34
3-(MeS)propyl isothiocyanate [505-79-3] Cauliflower 35
3-Butylphthalide [6066-49-5] Celery 35
Sedanolide [6415-59-4] Celery 35
(Z)-3-Hexenyl pyruvate [68133-76-6] Celery 4
Dimethyl sulfide [75-18-3] Corn 35
2-Acetyl-2-thiazoline [29926-41-8] Corn, fresh 35
(E,Z)-2,6-Nonadienal [557-48-2] Cucumber 32
(E)-2-Nonenal [2463-53-8] Cucumber 32
2-iso-butyl-3-methoxypyrazine [24683-00-9] Green bell pepper 34
1-Octen-3-ol [3391-86-4] Mushroom 34
1-Octen-3-one [4312-99-6] Mushroom 34
p-Mentha-1,3,8-triene [18368-95-1] Parsley 41
2-iso-Propyl-3-methoxypyrazine [25773-40-4] Pea (raw) 34
3-Methylthiopropanal [3268-49-3] Potato (boiled) 18
2-iso-Propyl-3-methoxypyrazine [25773-40-4] Potato (earthy) 34
2-Ethyl-6-vinylpyrazine [32736-90-6] Potato (baked) 34
2-Ethyl-3,5-dimethylpyrazine [13925-07-0] Potato chip 40
(Z)-3-Hexenal [6789-80-6] Tomatillo 36
(E,E )-2,4-Decadienal [25152-84-5] Tomatillo 36
2-iso-Butylthiazole [18640-74-9] Tomato (fresh) 18
(Z)-3-Hexenal [6789-80-6] Tomato (fresh) 31
of red beets and is detectable at a 100 ppt concentration. The flavor of raw peas
and peapods is attributable to 2-iso-propyl-3-methoxypyrazine (34).
The importance of C9 aldehydes to the character impact of cucumber flavor
was recently confirmed by calculating their odor unit values (ratio of concentra-
tion to odor threshold). (E,Z )-2,6-Nonadienal and (Z )-2-nonenal were determined
to be the principal odorants of cucumbers (32).
Tomatillo (Physalis ixocarpa Brot.) is a solanaceous fruit vegetable
similar in appearance to a small green tomato, which is used to prepare green
salsas in various Mexican dishes. The character impact compounds in toma-
tillo were recently established as (Z )-3-hexenal, (E,E )-2,4-decadienal, and

nonanal, which impart its dominant green flavor (36). Similar to tomato
flavor, -ionone and -damascenone provide modifying effects; however, toma-
tillo does not contain 2-iso-butylthiazole, a key character impact compound of
tomato.
Among the cruciform vegetables, cooked cabbage owes its dominant char-
acter impact flavor to dimethyl sulfide. In raw cabbage flavor, allyl isothiocya-
nate contributes sharp, pungent horseradish-like notes (35). Methyl methanethio-
sulfinate was observed to provide the character impact of sauerkraut flavor and
occurs in Brussels sprouts and cabbage (37,38). Compounds likely to be impor-
tant to the flavor of cooked broccoli include dimethyl sulfide and trisulfide,
nonanal, and erucin (4-[methylthio]butyl isothiocyanate) (34). Cooked cauli-
flower contains similar flavor components as broccoli, with the exception that 3-
(methylthio)propyl isothiocyanate is the characterizing thiocyanate (34). Rep-
resentative structures for vegetable flavor impact compounds are presented in
Figure 3.
Potato flavor is greatly influenced by methods of cooking or preparation.
Raw potato contains the characteristic earthy aroma component, 2-iso-propyl-
3-methoxypyrazine. A character impact compound common to boiled and baked
potatoes is methional (3-[methylthio]propanal). Baked potatoes contain Maillard
products such as 2-ethyl-3-methylpyrazine (earthy, nutty) and 2-ethyl-6-vinylpyr-
azine (buttery, baked potato) (34). In potato chips and French-fried potatoes, the
potato flavor character of methional is modified by volatile aromatics from frying
oils, such as (E,E )-2,4-decadienal, and thermally generated alkyl oxazoles pos-
sessing lactone-like flavors (34,39). The pyrazines 2-ethyl-3,5-dimethyl and 2,3-
diethyl-5-methylpyrazine are described as potato chip like (40).
Aroma impact compounds that are universal to all varieties of mushrooms
include 1-octen-3-ol (1 ppb threshold) and 1-octen-3-one (0.05 ppb threshold),
both of which have been described as having a fresh, wild-mushroom aroma (15).
However, 1-octen-3-one also possesses a metallic odor, particularly in the context
of oxidized oils (34).
The characteristic compound for raw carrot is 2-sec-butyl-3-methoxypyra-
zine, which has an extremely low (2 ppt) threshold value. Its sniffing port aroma
in gas chromatography-olfactometry has been described as raw carroty (34).
Unsaturated aldehydes contribute to the flavor of cooked carrot, the most signifi-
cant being (E )-2-nonenal (fatty-waxy) (34).
Compounds important to the aroma of celery include two lactones, 3-
butylphthalide and sedanolide (3-butyl-3a,4,5,6-tetrahydrophthalide), which pro-
vide impact at low concentration (3 ppm). Other terpene hydrocarbons such as
-selinene and limonene are present in greater abundance but do not significantly
contribute to celery flavor. A recent assessment of its odor threshold by Takeoka
388 McGorrin
FIGURE 3 Representative vegetable character impact flavor compounds.
suggests that 3-butylphthalide is the most significant character impact compound

for celery (35).
Key parsley aroma compounds were recently identified (41). The primary flavor
contributors were found to include p-mentha-1,3,8-triene (terpeny, parsleylike),
myrcene (metallic, herbaceous), 2-sec-butyl-3-methoxypyrazine (musty, earthy),
myristicin (spicy), linalool (coriander), (Z )-6-decenal (green, cucumber), and
(Z )-3-hexenal (green).
D. Maillard-Type and Brown Flavors

The flavor characteristics of heated sugar compounds, which possess caramel,
burnt sugar, and maple notes, include a family of structures that contain a methyl
enol-one group (42). As previously discussed, Furaneol (2,5-dimethyl-4-
hydroxy-(2H )-furan-3-one) has a sweet caramel, burnt sugar flavor with apprecia-
ble fruitiness, and occurs in beer, Arabica coffee, and white bread crust (43). Mal-
tol (3-hydroxy-2-methyl-4H-pyran-4-one) exhibits a sweet, burnt sugar, caramel
note similar to Furaneol, but not as strong or fruity. Ethyl maltol does not occur
in nature; however, it possesses a very intense sweet, caramel-like odor that is
four to six times more potent than maltol (11). Cyclotene (3-methyl-2-cyclopen-
ten-2-ol-1-one) has a strong maple-caramel flavor. 2-Methyltetrahydrofuran-3-
one has a very pleasant, sweet-caramel character (15). Sotolone (4,5-dimethyl-
3-hydroxy-2(5H )-furanone) was identified as a potent flavor impact compound
from raw cane sugar. At low concentrations, its aroma character is caramel-like
and elicits a powerful caramel, sweet, burnt sugar note, typical of unrefined cane
sugar (44). A summary of character impact compounds for thermally generated
flavors is outlined in Table 4.
An alternative source of characteristic heated flavor compounds is via the
Maillard pathway, the thermally induced reaction between amino acids and reduc-
ing sugars. Aroma constituents in chocolate, coffee, toasted bread, and popcorn
are products of Maillard reactions, in addition to flavors in roasted nuts and meats,
which are discussed in Sec. II.E and II.F. Guaiacols occur as pyrolysis products
of carbohydrates or lipids in smoked or char-broiled meats.
Furfurylthiol is the primary character impact compound for the aroma
of roasted Arabica coffee (45). It has a threshold of 5 ppt and smells like
freshly brewed coffee at concentrations between 0.01 and 0.5 ppb (46). At
higher concentrations it exhibits a stale coffee, sulfury note. Other potent odor-
ants in roasted coffee include 5-methylfurfurylthiol (0.05 ppb threshold), which
smells meaty at 0.51 ppb, and changes character to a sulfury mercaptan note
at higher levels (46). Furfuryl methyl disulfide has a sweet mocha coffee aroma
(15). A key aromatic that markedly contributes to coffee aroma is 3-mercapto-
3-methylbutyl formate. The pure compound has a blackcurrant-like, catty
note, however, in the context of brewed coffee, it contributes roast coffee like-
ness (47,48).
The principal impact aroma compounds of freshly prepared popcorn were
determined by Schieberle as 2-acetyltetrahydropyridine, 2-acetyl-1-pyrroline,
and 2-propionyl-1-pyrroline (49). The crackerlike aroma of the tetrahydropyri-
dine, which exists in two tautomeric forms (2-acetyl-1,4,5,6- and 2-acetyl-
3,4,5,6-tetrahydropyridine) was previously identified as the character compound
of Saltine crackers (50). The decrease of these compounds during storage was
directly correlated with staling flavor. Another popcornlike odor note, 2-acetyl-
390 McGorrin
TABLE 4 Character Impact Flavor Compounds in Cooked Flavors and Maillard-type

Systems (Chocolate, Coffee, Caramelized Sugar)
5-Methyl-2-phenyl-2-hexenal [21834-92-4] Chocolate 58
2-Methoxy-5-methylpyrazine [2882-22-6] Chocolate 4,59
iso-Amyl phenylacetate [102-19-2] Chocolate 4,59
2-Furfurythiol [98-02-2] Coffee 45
Furfuryl methyl disulfide [57500-00-2] Coffee (mocha) 15
Furaneol [3658-77-3] Fruity, burnt sugar 43
Sotolone [28664-35-9] Brown sugar 44
Phenylacetaldehyde [122-78-1] Honey 59
3-Methylbutanal [590-86-3] Malt 51
Maltol [118-71-8] Cotton candy 11
Ethyl maltol [4940-11-8] Caramel, sweet 11
2-Hydroxy-3-methyl-2-cyclopenten- [80-71-7] Maple 11
1-one
2-Acetyl-tetrahydropyridine [25343-57-1] Cracker (Saltine) 50
2-Acetyl-tetrahydropyridine [25343-57-1] Popcorn 49
2-Acetyl-1-pyrroline [85213-22-5] Popcorn 49
2-Propionyl-1-pyrroline [133447-37-7] Popcorn 49
2-Acetylpyrazine [22047-25-2] Popcorn 49
2-Acetyl-2-thiazoline [22926-41-8] Roasty, popcorn 56
5-Acetyl-2,3-dihydro-1,4-thiazine [164524-93-0] Roasty, popcorn 52
2-Acetyl-1-pyrroline [85213-22-5] Wheat bread crust 51
2-Acetyl-1-pyrroline [85213-22-5] Rice (Basmati) 51
2-ethyl-3,5-dimethylpyrazine [13925-07-0] Potato chip 56
2,3-diethyl-5-methylpyrazine [18138-04-0] Potato chip 56
2-Vinylpyrazine [4177-16-6] Roasted potato 56
(Z)-2-propenyl-3,5-(Me2 )pyrazine [55138-74-4] Roasted potato 56
Guaiacol [90-05-1] Smoky 4
4-Vinylguaiacol [7786-61-0] Smoky 15
pyrazine, was determined to make a minor contribution to the aroma of fresh

popcorn, because its odor intensity was considerably lower (51). An intense (0.06
ppt threshold) roasty, popcornlike odorant was recently identified in a ribose/
cysteine model system as 5-acetyl-2,3-dihydro-1,4-thiazine (52).
The flavor formed by the cooking of fragrant rice (e.g., Basmati) is de-
scribed as popcornlike, hence it is not surprising that 2-acetyl-1-pyrroline is the
character impact volatile (51). In masa corn tortillas, 2-aminoacetophenone pro-
vided the character impact resulting from the lime treatment of corn (53); in
corn chips, its contribution is modified by 2-acetyl-1-pyrroline and unsaturated
FIGURE 4 Representative thermally generated character impact flavor compounds.
aldehydes (54). Representative chemical structures for thermally generated flavor

impact compounds are shown in Figure 4.
Two compounds that create the characteristic odor notes in the pleasant
aroma of wheat bread crust have been identified as the popcornlike 2-acetyl-1-
pyrroline and 2-acetyltetrahydropyridine (51). The aroma of the bread crumb
portion is principally due to lipid-derived unsaturated aldehydes such as (E )-2-
nonenal and (E,E )-2,4-decadienal, which create stale aromas at high levels. The
malty notes that predominate in yeast and sourdough breads are attributed to 2-
and 3-methylbutanal and Furaneol (51,55).
392 McGorrin
Potato chip aroma is associated with 2-ethyl-3,5-dimethylpyrazine and

2,3-diethyl-5-methylpyrazine, whereas 2-vinylpyrazine and (Z )-2-propenyl-3,5-
dimethylpyrazine provide an intense roasted potato smell (56).
Chocolate represents a highly complex flavor system for which no single
character impact has been identified. Vanillin and Furaneol contribute to the
sweet, caramel background character of milk chocolate (57). 5-Methyl-2-phenyl-
2-hexenal provides a deep bitter, cocoa note, and is the aldol reaction product
from phenylacetaldehyde and 3-methylbutanal, two Strecker aldehydes formed
in chocolate (58). 2-Methoxy-5-methylpyrazine and isoamyl phenylacetate have
chocolate, cocoa, nutty and cocoa-like notes, respectively, and both are
used in synthetic chocolate flavors (59). Systematic studies of key odorants in
milk chocolate were performed using aroma extract dilution analysis; however,
character impact compounds unique to chocolate flavor were not reported (57,60).
E. Nut Flavors
Pyrazines are the major compound classes in peanuts, formed through the ther-
mally induced Maillard reaction (with the exception of methoxy pyrazines) (61).
Two pyrazines that represent peanut flavor character are 2,5-dimethylpyrazine
(nutty) and 2-methoxy-5-methylpyrazine (roasted nutty) (Table 5).
Benzaldehyde has long been known as the character impact of oil of bitter
almond. It possesses an intense almond-like flavor in the context of savory appli-
cations; in sweet systems, it becomes cherrylike. 5-Methyl-2-thiophene-carboxal-
dehyde also provides almond flavor character and occurs naturally in roasted
peanuts (59).
The character impact compound of hazelnuts, (E )-5-methyl-2-hepten-4-one
(filbertone), undergoes isomerization during the roasting process (62). Of the
four possible geometric and enantiomeric isomers formed, all exhibit the typical
TABLE 5 Character Impact Flavor Compounds in Nuts

Benzaldehyde [100-51-6] Almond 4

5-Methyl-2-thiophenecarboxal [13679-70-4] Almond 4,59
-Nonalactone [104-61-0] Coconut 61
-Decalactone [705-86-2] Coconut 61
Methyl(methylthio)pyrazine [21948-70-9] Hazelnut 59
(E )-5-Methyl-2-hepten-4-one [81925-81-7] Hazelnut 62
2,5-Dimethylpyrazine [123-32-0] Peanut 4
2-Methoxy-5-methylpyrazine [68358-13-5] Peanut 4
FIGURE 5 Representative nut character impact flavor compounds.
hazelnut aroma, but the trans-(S)-isomer has the strongest impact. Methyl(meth-
ylthio)pyrazine is a synthetic aroma chemical with the character of roasted al-
monds and hazelnuts (59). Structures are shown in Figure 5.
As previously discussed, the -lactones (e.g., -decalactone and -octalac-
tone) possess a coconut flavor character. However -nonalactone has the most
intense coconutlike aroma as an individual character impact compound, but it
occurs only in artificial coconut flavors (61). As the side-chain length increases,
the character of -lactones changes to peachlike (11).
F. Meat and Seafood Flavors

Sulfur-containing heterocyclic compounds are associated with meaty characteris-
tics. Two compounds with the most potent meaty impact include 2-methyl-3-
furanthiol (1 ppt) and the corresponding dimer, bis-(2-methyl-3-furyl) disulfide
(0.02 ppt) (18). Both substances have been identified in cooked beef and chicken
broth and have a strong meaty quality upon dilution. The disulfide has a recogniz-
able aroma character of rich aged-beef, prime-rib (15). Interestingly, both
compounds are produced from the thermal degradation of thiamin (63). A related
compound, 2-methyl-3-(methylthio)furan, is the character impact compound for
roast beef (18). Other potent modifiers, such as 2-acetyl-2-thiazoline, impart a
potent roasty, popcorn note, which enhances the meaty and roast flavor (64).
2-Ethyl-3,5-dimethylpyrazine and 2,3-diethyl-5-methylpyrazine also contribute
potent roasty notes to roast beef flavor (65). A summary of character impact
compounds for meat and seafood flavors is presented in Table 6.
394 McGorrin
TABLE 6 Character Impact Flavor Compounds in Meats and Fish

Dimethyl sulfide [75-18-3] Clam, oyster 69
4-Methylnonanoic acid [45019-28-1] Lamb 63
4-Methyloctanoic acid [54947-74-9] Lamb 63
2-Pentylpyridine [2294-76-0] Lamb 63
2-Acetyl-2-thiazoline [22926-41-8] Roasty (beef ) 64
4-Me-5-(2-hydroxyethyl)thiazole [137-00-8] Roasted meat 15
2-Ethyl-3,5-dimethylpyrazine [13925-07-0] Roasty (beef ) 65
2,3-diethyl-5-methylpyrazine [18138-04-0] Roasty (beef ) 65
2-Methyltetrahydrofuran-3-thiol [57124-87-5] Brothy, meaty 15
2-Methyl-3-furanthiol [28588-74-1] Meat, beef 18
Bis-(2-methyl-3-furyl) disulfide [28588-75-2] Aged, prime rib 18
12-Methyltridecanal [75853-49-5] Beef, stewed 67
(E,E)-2,4-Decadienal [25152-84-5] Chicken fat 15
Mercaptopropanone dimer Chicken broth 15
Pyrazineethanethiol [35250-53-4] Pork 15
(Z)-1,5-octadien-3-one [65767-22-8] Salmon, cod 71
(E,Z)-2,6-nonadienal [557-48-2] Trout, boiled 72
Pyrrolidino-2,4-(Me2)dithiazine [116505-60-3] Roasted shellfish 74
5,8,11-tetradecatrien-2-one [85421-52-9] Shrimp, cooked 73
2,4,6-Tribromophenol [118-79-6] Shrimp, ocean fish 75
A brothy compound associated with boiled beef, 4-methylthiazole-5-

ethanol (sulfurol), is a reaction flavor product from hydrolysis of vegetable
protein. It is suspected that a trace impurity (2-methyltetrahydrofuran-3-thiol) in
sulfurol is the actual beef broth character impact compound (15). Another
reaction product flavor chemical, mercaptopropanone dimer, has an intense
chicken-broth odor. A synthetic pyrazine, 2-pyrazineethanethiol, provides excel-
lent pork character (15).
Lipid components associated with meat fat, especially unsaturated alde-
hydes, play a significant role in species-characterization flavors. For example,
(E,Z )-2,4-decadienal exhibits the character impact of chicken fat and freshly
boiled chicken (66). (E,E )-2,6-Nonadienal has been suggested as the component
responsible for the tallowy flavor in beef and mutton fat (63). 12-Methyltridecanal
was identified as a species-specific odorant of stewed beef and provides a tallowy,
beeflike flavor character (67). Aldehydes provide desirable flavor character to
cooked meat, but they can contribute rancid and warmed-over flavors at high
concentrations, resulting from autoxidation of lipids (68).
Two fatty acids, 4-methyloctanoic and 4-methylnonanoic acid, provide the
characteristic flavor of mutton (63). 2-Pentylpyridine has been identified as the
most abundant alkylpyridine isolated from roasted lamb fat. This compound has
a fatty, tallowy aroma at an odor threshold of 0.6 ppb and is suspected to nega-
tively impact acceptance of lamb and mutton (63).
The fishy aroma of seafood is incorrectly attributed to trimethyl amine.
Flavor formation in fresh and saltwater fish results from complex enzymatic, oxida-
tive, and microbial reactions of n-3 polyunsaturated fatty acid precursors (e.g.,
eicosapentaenoic acid) (69,70). Hence, fish flavor is mostly composed of non-char-
acterizing planty or melon-like aromas from lipid-derived unsaturated car-
bonyl compounds. Examples are (Z)-1,5-octadien-3-one (geranium-like) in
boiled cod (71) and (E,Z)-2,6-nonadienal (cucumber-like) in boiled trout (72).
Three notable marine character-impact aroma exceptions are 5,8,11-
tetradecatrien-2-one, which exhibits a distinct seafood aroma character de-
scribed as cooked shrimp-like or minnow bucket (73). A second example
is an extremely potent odorant in cooked shellfish, including shrimp and clam,
identified by Kubota and coworkers (74) as pyrrolidino[1,2-e]-4H-2,4-dimethyl-
1,3,5-dithiazine. This dithiazine contributes a roasted character to boiled shell-
fish and has the lowest odor threshold recorded to date, 105 ppt in water. 2,4,6-
Tribromophenol and other bromophenol isomers have been associated with the
ocean-, brine-, and iodine-like flavor character in seafood such as Australian
ocean fish and prawns. The source of the bromophenols is thought to be poly-
chaete worms, which form an important part of the diet for many fish and prawn
species (75). Finally, dimethyl sulfide is the character aroma of stewed clams
and oysters (69). Representative structures for meat and seafood flavor impact
compounds are shown in Fig. 6.
G. Cheese and Dairy Flavors

With a few exceptions, many of the known important flavors in dairy products
do not provide characterizing roles. This is especially true for milk, cheddar
cheese, and cultured products, such as sour cream and yogurt. Delta-lactones
are important flavors in butter, buttermilk, and cheeses that are derived from
triglycerides containing hydroxyl fatty acids. Although not directly contributing
as dairy character impact compounds, lactones play key supporting roles. The
subject of key odor-active compounds in milk dairy flavors has been recently
reviewed (76,77). A summary of character impact compounds for cheese and
dairy flavors is presented in Table 7.
Urbach (78) recently discussed the formation of volatile flavor compounds
in different varieties of cheeses and provided a compilation of important aroma
compounds. A recent qualitative assessment by Sable and Cottenceau (79) surveys
the significant flavor volatiles that have been identified in soft mold-ripened cheeses,
including Camembert, Brie, blue, Gorgonzola, Muenster, and Limburger, among
others. Octen-3-ol, 2-phenylethanol, and 2-phenylethyl acetate are character impact
396 McGorrin
FIGURE 6 Representative meat and seafood character impact flavor compounds.
TABLE 7 Character Impact Flavor Compounds in Cheese and Dairy Products

2,3-Butanedione [431-03-8] Butter 81
-Decalactone [705-86-2] Butter 81
6-Dodecen--lactone [156318-46-6] Butter 81
2-Heptanone [110-43-0] Cheese, Blue 78
1-Octen-3-ol [3391-86-4] Cheese, Camembert 79
Butyric acid [107-92-6] Cheese, Cheddar 80
Methional [3268-49-3] Cheese, Cheddar 80
Skatole [83-34-1] Cheese, Cheddar 80
Homofuraneol [110516-60-4] Cheese, Swiss 82
Propionic acid [79-09-4] Cheese, Swiss 82
(Z)-4-Heptenal [6728-31-0] Cream 4
(E,E)-2,4-Nonadienal [5910-87-2] Cream 77
1-Nonen-3-one [24415-26-7] Milk 83
components in Camembert-type cheese; these compounds together with sulfur com-

pounds, 1-octen-3-one, and -decalactone are reported as the key aroma substances
for Camembert. 2-Heptanone, 2-nonanone and short- and moderate-chain fatty acids
are the dominant character compounds of blue cheese flavor. Sulfur compounds,
especially methanethiol, hydrogen sulfide, and dimethyl disulfide contribute to the
strong garlic/putrid aroma of soft-smear or surface-ripened cheeses.
By a wide margin, Cheddar is the most popular cheese flavor in North
America. Its flavor is described as sweet, buttery, aromatic, and walnut, yet
there is no general consensus among flavor chemists about the identity of individ-
ual compounds or groups of compounds responsible for Cheddar flavor. Reinec-
cius and Milo (80) concluded that butyric acid, acetic acid, methional, 2,3-
butanedione, and homofuraneol (5-ethyl-4-hydroxy-2-methyl-(2H)-furan-3-one)
are primary contributors to the pleasant mild flavor of Cheddar cheese. Important
contributors to Cheddar aroma are 2,3-butanedione, dimethyl sulfide/trisulfide,
and methanethiol (80).
The most aroma-active compounds in fresh sour cream butter were eluci-
dated as the character impact compound diacetyl, with supporting roles from
-decalactone, (Z )-6-dodeceno--lactone, and butyric acid (81).
A summary of potent flavor compounds in dairy products is presented in
Table 7. Representative structures of significant or newly identified flavorants, in-
FIGURE 7 Representative cheese and dairy character impact flavor compounds.

398 McGorrin
cluding homofuraneol in heated butter (76), mild cheddar cheese (80), and Swiss
cheese (82); -decalactone in butter and buttermilk (81); (Z)-6-dodecen--lactone
in butter (76) and 1-nonen-3-one in milk (83), are shown in Fig. 7.
III. CHARACTERIZING AROMAS IN OFF-FLAVORS
Flavor defects, so-called taints or off-flavors, are sensory attributes that are
not associated with the typical aroma and taste of foods and beverages. These
defects can range from subtle to highly apparent, and are often significant de-
tractors to food quality. Off-flavors can be produced due to several possible fac-
tors: contamination (air, water, packaging, or shipping materials); ingredient mis-
takes in processing; or generation (chemical or microbial) in the food itself. In
the latter instance, generation of off-flavors in foods may result from oxidation,
nonenzymatic browning, chemical reactions between food constituents, light-
induced reactions, or enzymatic pathways.
During the past 10 years, numerous complete and detailed reviews have
discussed the occurrence of off-flavors in food and packaging systems (8488).
The intent of this section is to summarize recent highlights and off-flavorants of
significance, without striving for comprehensiveness.
Lipid-derived volatile compounds play an important role in many food fla-
vors. These compounds contribute to the characteristic and desired note of a food
but can also cause off-flavors depending on their concentrations compared to
other sensorially relevant odorants. For example, the cardboard off-flavor of
butter oil is primarily related to (E )-2-nonenal, which is formed by the autoxida-
tion of palmitoleic acid (89). Carbonyl compounds formed by lipid peroxidation
were identified in cooked beef, which developed a warmed-over flavor from re-
heating after 2-day refrigerated storage (68). Warmed-over flavor was mainly
caused by formation of 4,5-epoxy-(E )-2-decenal (metallic) and hexanal
(green) notes, which were not present in freshly cooked beef. Similarly, for
boiled chicken, green cardboard-like metallic off-odors were formed during
refrigerated storage and reheating, primarily from a 7-fold increase in hexanal
(66). These and other unsaturated carbonyl compounds, including (E,E )-2,4-
decadienal (deep-fried) and 1-octen-3-one (mushroom-like/beany), and
the light-induced 3-methyl-2,4-nonanedione (strawy, lard-like, beany) are pri-
marily responsible for rancid off-flavors in soybean and canola oils through
oxidation of linoleic and linolenic acids (90,91). Melon odors are associated
with foods cooked in partially hydrogenated soybean oils that have undergone
oxidative deterioration during heating. 6-Nonenal (cucumber/melon) from
(Z,Z )-9,15-linoleic acid is reported to be a character flavor associated with
these deteriorated oils (92). (Z )-3-Hexenal and (Z,Z )-3,6-nonadienal were shown
to contribute substantially to the fatty, fishy off-flavor of boiled trout that was
in frozen storage for several months before cooking (72). Other carbonyl com-
pounds that are likely to contribute to characteristic fishy off-flavors in oxi-
dized seafoods because of their low odor thresholds include 2,6-nonadienal and
1,5-octadien-3-one (69). A summary of off-flavor impact compounds for foods,
beverages, and packaging materials is presented in Table 8, with representative
chemical structures in Fig. 8.
trans-2-Nonenal is considered to be the characteristic volatile responsible
for a stale, cardboard off-flavor in aged packaged beer. Mechanistic studies
using labeled nonenal confirmed that cardboard off-flavor in finished beer arises
from lipid autoxidation during wort boiling, and not from lipoxygenase activity
during the mashing step (93). Studies also revealed that 2-furfuryl ethyl ether is
responsible for an astringent, stale off-flavor in beer (94). Stale flavor was repro-
duced by adding the furfuryl ether and trans-2-nonenal to fresh beer, but not when
either compound was added individually. Lipid-derived (E,E )-2,4-decadienal, in
addition to hexanal, (E )-2-octenal, and (E )-2-nonenal, were shown to be the most
potent off-flavor compounds in precooked vacuum-packaged potatoes (95). In
dry raw spinach, (Z )-1,5-octadien-3-one and methional are responsible for a
fishy off-flavor at a 1 : 100 ratio. 3-Methyl-2,4-nonanedione produces a hay-
like off-flavor character from oxidative degradation of furan fatty acids in dry
spinach (96), as well as in dry parsley (97).
Exposure of beer to light has been shown to produce 3-methyl-2-butene-
1-thiol, which produces a skunky off-flavor in sun-struck or light-struck
ales (98,99). This mercaptan has a sensory threshold of 0.05 ppb in beer. It results
from complex photo-induced degradations of isohumulones (hop-derived, bitter
iso-acids) to form free-radical intermediates, which subsequently react with the
thiol group of cysteine. Lightstruck off-flavor can be controlled in beer through
packaging technology (colored glass bottles), use of chemically modified hop
bitter acids, antioxidants, or its precipitation with high molecular weight gallotan-
nins and addition of zinc salts (100). In addition to dimethyl sulfide, thioesters
have been reported to contribute a cabbagy, rubbery off-note that sometimes oc-
curs in beer, the most significant being S-methyl hexanethioate, which has a de-
tection threshold of 1 ppb (16a). Diacetyl can produce an undesirable buttery
off-character in beer through accelerated fermentation, whereby brewers yeast
has not converted all of the diketone intermediates to flavor-inactive acetoin and
2,3-butanediol (98).
Strecker aldehydes are a frequent source of off-flavors in fermented prod-
ucts. Development of off-flavors in oxidized white wines typically marks the end
of shelf life. Methional (3-methylthiopropionaldehyde) was identified as produc-
ing a cooked vegetables off-flavor character in a young white wine that had
undergone spontaneous oxidation (101). Methional levels increased in wines
spiked with methionol or methionine, suggesting its formation via direct peroxi-
dation or Strecker degradation of methionine. Methional was recently demon-
400
TABLE 8 Off-Flavor Impact Compounds in Food and Beverage Products
Impact compound(s) CAS registry no. Off-flavor Occurrence Reference
Geosmin [19700-21-1] Musty, earthy Catfish, wheat, water 107,110

2,4,6-Trichloroanisole [87-40-1] Musty, moldy Coffee, wine corks 47,110
2-Methylisoborneol [2371-42-8] Earthy, musty Coffee, catfish, beans 47,107,
110
4,4,6-Trimethyl-1,3-dioxane [1123-07-5] Musty Packaging film 113
Iodocresol [16188-57-1] Medicinal Lemon cake mix 112
(E)-2-Nonenal [2463-53-8] Cardboard, stale Beer, packaged 98
Cardboard Oxidized butter 89
Cardboard Soybean lecithin 90
trans-4,5-Epoxy-(E)-2-decenal [134454-31-2] Metallic Warmed-over (beef ) 68
Metallic Oxidized soybean oil 90
Metallic Butter 43
(E,Z)-2,6-nonadienol [7786-44-9] Metallic Buttermilk 121
1-Octen-3-one [4312-99-6] Metallic, mushroom Butterfat 114
Beany Rancid soybean oil 90,91
Hexanal [66-25-1] Green grass Rancid soybean oil 90,91
Rancid Warmed-over (beef ) 68
(E,E)-2,4-Decadienal [25152-84-5] Deep-fried Rancid soybean oil 90
2,3-Diethyl-5-methylpyrazine [18138-04-0] Roasty, earthy Soybean lecithin 90
6-Nonenal [6728-35-4] Cucumber, melon Part.-hydrog. soy oil 92
McGorrin
(Z)-3-Hexenal [6789-80-6] Fatty, fishy Aged trout 72
(Z,Z)-3,6-Nonadienal [21944-83-2] Fatty, fishy Aged trout 72
5-Androst-16-en-3-one [18339-16-7] Urine Boar meat 86,125
Character Impact Compounds
Skatole [83-34-1] Fecal-like Boar meat 125
Fecal Potato chips 86
Medicinal Beef 126
2-Furfuryl ethyl ether [6270-56-0] Stale, astringent Beer 94
3-Methyl-2-butene-1-thiol [5287-45-6] Skunky Beer (light-struck) 98,99
S-Methyl hexanethioate [2432-77-1] Cabbagy, rubbery Beer 16a
Methional [3268-49-3] Worty Beer (alcohol-free) 102
Cooked vegetables Oxidized white wine 96
Sunlight off-flavor Milk 87,118
(E)-1,3-Pentadiene [504-60-9] Kerosene Cheese (sorbic acid) 86,122
Tetradecanal [124-25-4] Sickening, aldehydic Dried milk powder 123
-Ionone [79-77-6] Hay-like Dried milk powder 123
Benzothiazole [95-16-9] Sulfuric, quinoline Dried milk powder 123
2-Aminoacetophenone [551-93-9] Gluey, glutinous Milk powder, casein 115,116
White wine 117
3-Methyl-2,4-nonanedione [113486-29-6] Strawy, beany Soy oil (light-induced) 91
Hay-like Dried spinach, parsley 96,97
(Z)-1,5-Octadien-3-one [65767-22-8] Fishy Dry spinach, old fish 69,96
Sotolone [28664-35-9] Burnt, spicy Citrus soft drink 119
4-Vinylguaiacol [7786-61-0] Orange juice 128
Apple juice 129
Beer, wort 130
Bis(2-methyl-3-furyl)disulfide [28588-75-2] Vitamin B1 odor Thiamin degradation 120
4-Methyl-2-isopropylthiazole [15679-13-7] Vitamin, cabbage Orange juice (Vit B2 ) 104
401
402 McGorrin
FIGURE 8 Representative off-flavor character impact compounds.

strated to impart a worty off-flavor in alcohol-free beer, with more sensory

significance than was previously attributed to 3-methyl- and 2-methylbutanal for
this off-taste (102). In an alcohol-free beer medium, perception of worty off-
flavors is strengthened by the absence of ethanol and by the higher level of sugars.
Finally, methional is responsible for sunlight flavor in photooxidized milk,
resulting from the oxidation of methionine in the presence of riboflavin as a
sensitizer (103). Another vitamin-derived off-odor problem was recently de-
scribed in which a pineapple fruit juice beverage was fortified with riboflavin.
The vitamin, cabbage, brothy, vegetable soup off-odor was characterized as
4-methyl-2-isopropylthiazole, which resulted from riboflavin-sensitized Strecker
degradation of valine, cysteine, and methionine, followed by reaction of the re-
sulting aldehydes with ammonia and hydrogen sulfide (104).
Sulfur compounds present in wine can have a detrimental effect on aroma
character, producing odors described as garlic, onion and cauliflower, so-
called Boeckser aroma. This sulfurous character is correlated with with 2-methyl-
3-hydroxythiophene, 2-methyl-3-furanthiol and ethanethiol, and their con-
centrations in wine are influenced by winery procedures and the use of certain
winemaking yeasts (105). Off-flavors in European wines were associated with
the nonvolatile bis(2-hydroxyethyl) disulfide, a precursor to the poultry-like
character of 2-mercaptoethanol and hydrogen sulfide (106).
Musty aromas and flavors are a major problem in a variety of foods and
packaging materials, and many of the causative character impact materials have
powerful impact (107). For example, 2,4,6-trichoroanisole is a highly odorous
metabolite of a fungus that attacks wood, paperboard packaging, and wine corks
(85,86). Trichloroanisole has a musty, haylike odor and possesses an extremely
low odor threshold (0.05 ppt). Both choloroanisoles and their corresponding
phenols can provide musty off-flavor character to foods and process water. For
example, chloroanisoles were related to musty off-flavor of raisins (108). About
20% of Brazilian coffee production forms the so-called Rioy defect character-
ized by a strong off-flavor that is often described as medicinal, phenolic, or
iodine-like (109). Occasionally, this defect also occurs in coffees from other
origins. 2,4,6-Trichloroanisole in concentrations ranging from 1100 ppb was
identified as the most likely key compound for the Rio off-flavor as analyzed
by capillary GC, GC-sniffing, and GC-MS. 2,4,6-Trichlorophenol, the probable
precursor, was also found in most of these samples. Adding trichloroanisole to
freshly brewed coffee imparted to it the same off-flavor notes as described in
actual Rio coffee. Its perception threshold in coffee brew was found to be 8 ppt
by aroma and 12 ppt for taste (109). In Robusta coffees, 2-methylisoborneol
provides an earthy, tarry character at 5 ppt, and must be removed during pro-
cessing to approach the flavor of Arabica coffees (47). Geosmin (1,10-dimethyl-
9-decalol) imparts an earthy-musty off-odor to drinking water, and a muddy
odor to catfish and tilapia fish that live in brackish water (107,110,111). It is
404 McGorrin
produced from Actinomycetes microorganisms in soil, planktonic algae, and

fungi. 2-Methylisoborneol also provides an earthy-musty character to catfish
(110). A medicinal off flavor was produced in a lemon-flavored cake mix by
reaction between two minor ingredients, p-cresol from the lemon flavor, and io-
dine from iodized salt. The resulting iodocresol was shown to possess a medici-
nal character at 0.2 ppb aroma threshold (112). We studied a musty off-
odor in printed plastic film, which was not a phenol as expected but was identified
as 4,4,6-trimethyl-1,3-dioxane (113). It was formed during film manufacture as
a reaction product between 2-methyl-2,4-pentanediol, a solvent coating to facili-
tate ink adhesion, and formaldehyde, a component in the ink.
The same compounds that have positive character impact flavor at low
levels can become off-flavors when at higher concentrations or in different food
contexts. For example 1-octen-3-one (mushroom) becomes metallic in dairy
products and oxidized vegetable oils (114). Indole and skatole in Cheddar cheese
flavor becomes fecal in the context of potato chips (86). 2-Aminoacetophenone
(foxy character) in Concord grape imparts a gluey flavor in milk powder
and casein through degradation of tryptophan (115,116). In white wines, such
as Riesling, it provides an untypical ageing note from either tryptophan or
kynurenine precursors (117). Sunlight off-flavor in milk (cardboard-like) can
result from photooxidation of milk exposed to high intensity fluorescent light or
sunlight, which converts methionine to methional, a flavor impact compound in
boiled potatoes. Methional is further degraded to other impact sulfur compounds
including dimethyl disulfide (118). Sotolone, the furanone character impact flavor
of fenugreek (Table 1), has recently been shown to cause a burnt, spicy off-
flavor in citrus soft drinks, generated by reaction of ascorbic acid with ethanol
(119). Whereas 2-isopropyl-3-methoxy pyrazine contributes to pea and earthy-
potato flavor character, it provides an undesirable peasy off-flavor to Ruanda
coffee (47) and fermented cocoa beans (86). Bis(2-methyl-3-furyl)disulfide,
which contributes an aged, prime rib flavor in beef, is the principal B-vitamin
off odor resulting from thiamin degradation (120).
Off-flavors in dairy products have been reviewed (86,87), but there are
several recent developments. In sour-cream buttermilk, the key odorant responsi-
ble for a metallic off-flavor was identified as (E,Z )-2,6-nonadienol (120). During
cream fermentation, its formation occurs from peroxidation of alpha-linolenic
acid to generate the 2,6-nonadienal precursor, with subsequent reduction to the
dienol by starter culture reductases that remain active during storage. Metallic
off-flavors are not formed readily in fermented sweet-cream buttermilk due to
the much lower concentration of alpha-linolenic acid present (121). The common
use of sorbic acid and potassium sorbate as a mold inhibitor in commercial dairy
products often produces an off-flavor described as kerosene, plastic-, or paint-
like, which may incorrectly be attributed to packaging materials. The flavorant
is (E )-1,3-pentadiene and results from decarboxylation of sorbates by lactic acid

bacteria in yogurt, cheese, and margarine (86,122).
Other significant recurring themes include studies on the characteristic
off-flavor in spray-dried skim milk powder, which was related to tetradecanal
(sickening, aldehydic), -ionone (hay-like), and benzothiazole (sulfuric,
quinoline) contributions at low-ppb levels (123). Fruity (pineapple-like)
off-flavors in pasteurized milk indicate the presence of high levels of ethyl bu-
tyrate, -hexanoate, -octanoate, and -decanoate esters; rancid, soapy tastes arise
from decanoic and dodecanoic acids (124).
As previously discussed, lipid oxidation is generally related to the flavor
deterioration in meat and meat products. However, boar taint, an intense urine-
like off-odor, is attributed primarily to the flavor synergy between two com-
pounds in boar fat, androstenone (5--androst-16-ene-3-one) from testes and the
fecal-like skatole (3-methyl indole) from tryptophan breakdown (125). Gener-
ally, women tend to be more sensitive to the odor than men. Skatole has also
been implicated in a medicinal off-odor in beef (126).
Off-flavors in citrus oils such as orange oil have been related to autoxidation
of limonene to (S)-carvone (caraway, Fig. 1) and carveol, producing woody,
turpeny off-flavors (127). Other off-flavors in orange juice arise from 4-vinyl-
guaiacol, which contributes an old fruit and rotten character, due to degra-
dation of ferulic acid (128). Conversion of ferulic acid to 4-vinylguaiacol by yeast
contaminants, with corresponding off-flavor development, has been reported in
unpasteurized apple juice (129) and in beers and worts (130). Thermal abuse
during processing or elevated temperature storage of orange and grapefruit juice
produces Furaneol, which contributes a sweet, pineapple defect (128).
IV. CONCLUSION
The objective of this chapter was to provide a current review of character impact
compounds in flavors and off-flavors. Particular emphasis was placed on com-
pounds that have been identified in natural flavor systems. The summarized data
can be applied in creative flavor compounding efforts to replicate and monitor
production of flavors and in food processing to ensure flavor quality.
REFERENCES
1. Volatile Compounds in Food. Qualitative and Quantitative Data, Seventh ed.

(L. M. Nijssen, C. A. Visscher, H. Maarse, L. C. Willemsens and M. H. Boelens,
eds.), TNO Biotechnology and Chemistry Institute, Zeist, The Netherlands, 1996.
406 McGorrin
2. S. S. Chang, Food flavors, Food Technology 43 (12):99 (1989).

3. R. Emberger, An analytical approach to flavor research, Cereal Foods World, 30
(10):691 (1985).
4. F. Fischetti, Flavors and spices, in: Kirk-Othmer Encyclopedia of Chemical Tech-
nology, Fourth ed., Vol. 11 (M. Howe-Grant, ed.), 1994, p. 16.
5. P. Darriet, T. Tominaga, V. Lavigne, J.-N. Biodron, and D. Dubourdieu, Identifica-
tion of a powerful aromatic component of Vitis vinifera L. var. Sauvignon wines:
4-mercapto-4-methylpentan-2-one, Flav. Fragr. J. 10:385 (1995).
6. H. Maarse in Volatile Compounds in Foods and Beverages (H. Maarse, ed.) Marcel
Dekker, New York, NY, 1991, p. 1.
pounds and their precursors, in: Characterization of Food: Emerging Methods
(A. G. Gaonkar, ed.), Elsevier, New York, NY , 1995, p. 403.
8. B. H. Mistry, T. Reineccius, and L. K. Olson, Gas chromatography-olfactometry
Aroma (R. Marsili, ed.), Marcel Dekker, New York, NY, 1997, p. 265.
9. W. Pickenhagen, Flavor chemistrythe last 30 years, in: Flavor Chemistry: Thirty
Years of Progress (R. Teranishi, E. L. Wick, I. Hornstein, eds.), Kluwer Academic/
Plenum, New York, NY, 1999, p. 75.
10. M. H. Boelens, Spices and condiments II, in: Volatile Compounds in Foods and
Beverages, (H. Maarse, ed.), Marcel Dekker, New York, NY, 1991, p. 449.
11. K. Bauer, D. Garbe, and H. Surburg, Common Fragrance and Flavor Materials,
Third ed., Wiley-VCH, Weinheim, Germany, 1997.
12. I. Blank, J. Lin, S. Devaud, R. Fumeaux, and L. B. Fay, The principal flavor compo-
nents of fenugreek (Trigonella foenum-graecum L.) in: Spices: Flavor Chemistry
and Antioxidant Properties (S. J. Risch and C.-T. Ho, eds.) ACS Symposium Series
13. R. D. Hiserodt, C.-T. Ho, and R. T. Rosen, The characterization of volatile and
semivolatile components in powdered turmeric by direct thermal extraction gas
chromatographymass spectrometry, in: Spices: Flavor Chemistry and Antioxi-
dant Properties (S. J. Risch and C.-T. Ho, eds.), ACS Symposium Series 660, Amer-
ican Chemical Society, Washington, DC, 1997, p. 80.
14. K. R. Cadwallader, H. H. Baek, and M. Cai, Characterization of saffron flavor
by aroma extract dilution analysis in Spices: Flavor Chemistry and Antioxidant
Properties (S. J. Risch and C.-T. Ho, eds.), ACS Symposium Series 660, American
Chemical Society, Washington, DC, 1997, p. 66.
15. D. Rowe, More fizz for your buck: high-impact aroma chemicals, Perfumer & Fla-
vorist 25:(9/10)1 (2000).
16. a). G. Ohloff, I. Flament, and W. Pickenhagen, Flavor chemistry, Food Rev. Intl.
1:99 (1985); b). T. Jagella and W. Grosch, Flavor and off-flavor compounds of
black and white pepper (Piper nigrum L.). II. Odor activity values of desirable and
undesirable odorants of black pepper, Eur. Food Res. Technol. 209:22 (1999).
17. I. Blank, A. Sen, and W. Grosch, Sensory study on the character-impact flavor
compounds of dill herb (Anethum graveolens L.), Food Chemistry, 43:337 (1992).
18. M. H. Boelens and L. J. van Gemert, Volatile character-impact sulfur compounds
and their sensory properties, Perfumer & Flavorist, 18 (3): 29 (1993).
19. W. M. Randle, Onion flavor chemistry and factors influencing flavor intensity, in:
Spices: Flavor Chemistry and Antioxidant Properties (S. J. Risch and C.-T. Ho,
eds.), ACS Symposium Series 660, American Chemical Society, Washington, DC,
1997, p. 41.
20. E. Block, S. Naganathan, D. Putman, and S.-H. Zhao, Allium chemistry: HPLC
analysis of thiosulfinates from onion, garlic, wild garlic (Ramsoms), leek, scallion,
shallot, elephant (great-headed) garlic, chive, and Chinese chive, J. Agric Food
Chem., 40:2418 (1992).
21. S. Widder, C. S. Luntzel, T. Dittner, and W. Pickenhagen, 3-Mercapto-2-methyl-
pentan-1-ol, a new powerful aroma compound, J. Agric. Food Chem. 48:418
(2000).
22. R. C. Berger, Fruits I, in: Volatile Compounds in Foods and Beverages, (H. Maarse,
ed.), Marcel Dekker, New York, NY, 1991, p. 283.
23. a). D. D. Roberts and T. E. Acree, Developments in the isolation and characteriza-
tion of -damascenone precursors from apples, in: Fruit Flavors: Biogenesis, Char-
acterization, and Authentication (R. L. Rouseff and M. M. Leahy, eds.), ACS Sym-
posium Series 596, American Chemical Society, Washington, DC, 1995, p. 190;
b). A. C. Noble, D. L. Elliott-Fisk, and M. S. Allen, Vegetative flavor and methoxy-
pyrazines in Cabernet Sauvignon, ibid, p. 226.
24. M. Larsen, L. Poll, and C. E. Olsen, Evaluation of the aroma composition of some
strawberry (Fragaria ananassa Duch) cultivars by use of odor threshold values,
Z. Lebensm. Unters. Forsc. 195:536 (1992).
25. H. H. Baek and K. R. Cadwallader, Contribution of free and glycosidically bound
volatile compounds to the aroma of muscadine grape juice, J. Food Sci. 64:441
(1999).
26. P. Schieberle and T. Hofmann, Evaluation of the character impact odorants in fresh
strawberry juice by quantitative measurements and sensory studies on model mix-
tures, J. Agric. Food Chem. 45:227 (1997).
27. M. G. Kolor, Identification of an important new flavor compound in Concord
grapes: Ethyl 3-mercaptopropionate, J. Agric. Food Chem. 31:1125 (1983).
28. E. Demole, P. Enggist, and G. Ohloff, 1-p-Menthene-8-thiol: a powerful flavor
impact constituent of grapefruit juice (Citrus paradisi MacFayden), Helv. Chim.
Act 65:1785 (1982).
29. P. E. Shaw, Fruits II, in: Volatile Compounds in Foods and Beverages (H. Maarse,
30. M. Larsen and L. Poll, Odor thresholds of some important aroma compounds in
raspberries, Z. Lebensm. Unters. Forsch. 191:129 (1990).
31. R. G. Buttery, Quantitative and sensory aspects of flavor of tomato and other vege-
tables and fruits, in: Flavor Science: Sensible Principles and Techniques (T. E.
Acree and R. Teranishi, eds.), ACS Books, American Chemical Society, Washing-
ton, DC, 1993, p. 259.
32. P. Schieberle, S. Ofner, and W. Grosch, Evaluation of potent odorants in cucumbers
(Cucumis sativus) and muskmelons (Cucumis melo) by aroma extract dilution anal-
ysis, J. Food Sci. 55:193 (1990).
33. M. Guntert, G. Krammer, H. Sommer, and P. Werkhoff, The importance of the
vacuum headspace method for the analysis of fruit flavors, in: Flavor Analysis:
408 McGorrin
Developments in Isolation and Characterization (C. J. Mussinan and M. J. Mor-

ello, eds.), ACS Books, American Chemical Society, Washington, DC, 1998,
p. 38.
34. F. B. Whitfield and J. H. Last, Vegetables in: Volatile Compounds in Foods and
Beverages (H. Maarse, ed.), Marcel Dekker, New York, NY, 1991, p. 203.
35. G. Takeoka, Flavor chemistry of vegetables, in: Flavor Chemistry: Thirty Years
of Progress (R. Teranishi, E. L. Wick, I. Hornstein, eds.), Kluwer Academic/Ple-
num, New York, NY, 1999, p. 287.
36. R. J. McGorrin and L. Gimelfarb, Comparison of flavor components in fresh and
cooked tomatillo with red plum tomato, in: Food Flavors: Formation, Analysis, and
Packaging Influences (E. T. Contis, C.-T. Ho, C. J. Mussinan, T. H. Parliment, F.
Shahidi, and A. M. Spanier, eds.), Elsevier, New York, NY, Developments in Food
Science Vol. 40, 1998, p. 295.
37. H. S. Marks, J. A. Hilson, H. C. Leichtweis, and G. S. Stoewsand, S-Methylcysteine
sulfoxide in Brassica vegetables and formation of methyl methanethiosulfinate
from Brussels sprouts, J. Agric. Food Chem. 40:2098 (1992).
38. H.-W. Chin and R. C. Lindsay, Mechanisms of formation of volatile sulfur com-
pounds following the action of cysteine sulfoxide lyases, J. Agric. Food Chem. 42:
1529 (1994).
39. R. K. Wagner and W. Grosch, Key odorants of French fries, J. Amer. Oil Chem.
Soc. 75:1385 (1998).
40. P. Schnermann and P. Schieberle, Evaluation of key odorants in milk chocolate
and cocoa mass by aroma extract dilution analyses, J. Agric. Food Chem. 45:867
(1997).
41. C. Masanetz and W. Grosch, Key odorants of parsley leaves (Petroselinum cris-
pum [Mill.] Nym. ssp. crispum) by odor-activity values, Flavor Fragr. J. 13:115
(1998).
42. R. Scarpellino and R. J. Soukup, Key flavors from heat reactions of food ingredi-
ents, in: Flavor Science: Sensible Principles and Techniques (T. E. Acree and R.
Teranishi, eds.) ACS Books, American Chemical Society, Washington, DC, 1993,
p. 309.
pounds and their precursors, in: Characterization of Food: Emerging Methods
(A. G. Gaonkar, ed.), Elsevier Science, New York, 1995, p. 403.
44. A. Kobayashi, Sotolon: identification, formation, and effect on flavor, in: Flavor
Chemistry: Trends and Developments (R. Teranishi, R. G. Buttery, and F. Shahidi,
eds.), ACS Symposium Series 388, American Chemical Society, Washington, DC,
1989, p. 49.
45. W. Grosch, Flavor of coffee. A review, Nahrung 42:344 (1998).
46. I. Flament, Coffee, cocoa and tea, in: Volatile Compounds in Foods and Beverages,
(H. Maarse, ed.), Marcel Dekker, New York, NY, 1991, p. 617
47. O. G. Vitzthum, Thirty years of coffee chemistry research, in: Flavor Chemistry:
Thirty Years of Progress (R. Teranishi, E. L. Wick, I. Hornstein, eds.), Kluwer
Academic/Plenum, New York, NY, 1999, p. 117.
48. P. Semmelroch and W. Grosch, Studies on character impact odorants of coffee
brews, J. Agric. Food Chem. 44:537 (1996).
49. P. Schieberle, Quantitation of important roast-smelling odorants in popcorn by sta-

ble isotope dilution assays and model studies on flavor formation during popping,
J. Agric. Food Chem. 43:2442 (1995).
50. P. Schieberle and W. Grosch, Bread flavor, in: Thermal Generation of Aromas
(T. H. Parliament, R. J. McGorrin, and C.-T. Ho, eds.), ACS Symposium Series
51. W. Grosch and P. Schieberle, Flavor of cereal productsa review, Cereal Chem.
74:91 (1997).
52. T. Hofmann, R. Hassner, and P. Schieberle, Determination of the chemical structure
of the intense roasty, popcorn-like odorant 5-acetyl-2,3-dihydro-1,4-thiazine, J.
Agric. Food Chem. 43:2195 (1995).
53. R. G. Buttery and L. C. Ling, Importance of 2-aminoacetophenone to the flavor
of masa corn flour products, J. Agric. Food Chem. 42:1 (1994).
54. R. G. Buttery and L. C. Ling, Additional studies on flavor components of corn
tortilla chips, J. Agric. Food Chem. 46:2764 (1998).
55. G. Zehentbauer and W. Grosch, Crust aroma of baguettes. I. Key odorants of ba-
guettes prepared in two different ways, J. Cereal Sci. 28:81 (1998).
56. T. Hofmann and P. Schieberle, Identification of key aroma compounds generated
from cysteine and carbohydrates under roasting conditions, Z. Lebensm. Unters.
Forsch. 207:229 (1998).
57. P. Schieberle and P. Pfnuer, Characterization of key odorants in chocolate, in: Fla-
vor Chemistry: Thirty Years of Progress (R. Teranishi, E. L. Wick, I. Hornstein,
eds.), Kluwer Academic/Plenum, New York, NY, 1999, p. 147.
58. M. van Praag, H. S. Stein, and M. S. Tibetts, Steam volatile constituents of roasted
cocoa beans, J. Agric. Food Chem. 16:1005 (1968).
59. G. A. Burdock, ed., Fenarolis Handbook of Flavor Ingredients, 3rd ed., Vol. 2,
CRC Press, Boca Raton, FL, 1995.
60. P. Schnermann and P. Schieberle, Evaluation of key odorants in milk chocolate
and cocoa mass by aroma extract dilution analyses, J. Agric. Food Chem. 45:867
(1997).
61. J. Maga, Nuts, in: Volatile Compounds in Foods and Beverages (H. Maarse, ed.),
Marcel Dekker, New York, NY, 1991, p. 671.
62. M. Guentert, R. Emberger, R. Hopp M. Koepsel, W. Silberzahn, and P. Werkhoff,
Chirospecific analysis in flavor and essential oil chemistry. A. Filbertonethe
character impact compound of hazel-nuts, Z. Lebensm. Unters. Forsch. 192:108
(1991).
63. D. S. Mottram, Meat, in: Volatile Compounds in Foods and Beverages (H. Maarse,
64. C. Cerny and W. Grosch, Quantification of character-impact odor compounds of
roasted beef, Z. Lebensm. Unters. Forsch. 196:417 (1993).
65. C. Cerny and W. Grosch, Precursors of ethyldimethylpyrazine isomers and 2,3-
diethyl-5-methylpyrazine formed in roasted beef, Z. Lebensm. Unters. Forsch. 198:
210 (1994).
66. J. Kerler and W. Grosch, Character impact odorants of boiled chicken: changes
during refrigerated storage and reheating, Z. Lebensm. Unters. Forsch. 205:232
(1997).
410 McGorrin
67. H. Guth and W. Grosch, 12-Methyltridecanal, a species-specific odorant of stewed

beef, Lebensm. Wiss. und Technol. 26:171 (1993).
68. J. Kerler and W. Grosch, Odorants contributing to warmed-over flavor (WOF) of
refrigerated cooked beef, J. Food Sci. 61:1271 (1996).
69. D. B. Josephson, Seafood, in: Volatile Compounds in Foods and Beverages (H.
Maarse, ed.), Marcel Dekker, New York, NY, 1991, p. 179.
70. R. J. Gordon and D. B. Josephson, Surimi seafood flavors, in: Surimi and Surimi
Seafood (J. W. Park, ed.), Marcel Dekker, NY, 2000, p. 393.
71. C. Milo and W. Grosch, Changes in the odorants of boiled salmon and cod as
affected by the storage of the raw material, J. Agric. Food Chem. 44:2366 (1996).
72. C. Milo and W. Grosch, Changes in the odorants of boiled trout (Salmo fario) as
affected by the storage of the raw material, J. Agric. Food Chem. 41:2076 (1993).
73. C. Karahadian and R. C. Lindsay, Evaluation of compounds characterizing fishy
flavors in fish oils, J. Am. Oil Chem. Soc. 66:953 (1989). A. Kobayashi, K. Kubota,
M. Iwamoto, and H. Tamura, Syntheses and sensory characterization of 5,8,11-
tetradeca-trien-2-one isomers, J. Agric. Food Chem. 37:151 (1989)
74. K. Kubota, A. Nakamoto, M. Moriguchi, A. Kobayashi, and H. Ishii, Formation of
Pyrrolidino[1,2-e]-4H-2,4-dimethyl-1,3,5-dithiazine in the volatiles of boiled short-
necked clam, clam, and corbicula, J. Agric. Food Chem. 39:1127 (1991).
75. F. B. Whitfield, M. Drew, F. Helidoniotis, and D. Svoronos, Distribution of bromo-
phenols in species of marine polychaetes and bryozoans from eastern Australia and
the role of such animals in the flavor of edible ocean fish and prawns (shrimp), J.
Agric. Food Chem. 47: 4756 (1999).
76. T. H. Parliment and R. J. McGorrin, Critical Flavor Compounds in Dairy Products,
in: Flavor Chemistry: Industrial and Academic Research, S. J. Risch, C.-T. Ho,
Eds., ACS Symposium Series 756, ACS Books, Washington, D.C., 2000, p. 44.
77. L. Schutte, Development and application of dairy flavors, in: Flavor Chemistry:
Thirty Years of Progress (R. Teranishi, E. L. Wick, I. Hornstein, eds.), Kluwer
Academic/Plenum, New York, NY, 1999, p. 155.
78. G. Urbach, The flavor of milk and dairy products: II. Cheese: contribution of vola-
tile compounds, Int. J. Dairy Technol. 50:79 (1997).
79. S. Sable and G. Cottenceau, Current knowledge of soft cheeses flavor and related
compounds, J. Agric. Food Chem. 47:4825 (1999).
80. C. Milo and G. A. Reineccius, Identification and quantification of potent odorants in
regular-fat and low-fat mild cheddar cheese, J. Agric. Food Chem. 45:3590 (1997).
81. P. Schieberle, K. Gassenmeier, H. Guth, A. Sen, and W. Grosch, Character impact
odor compounds of different kinds of butter, Lebensm. Wiss. und Technol. 26:347
(1993).
82. M. Preininger and W. Grosch, Evaluation of key odorants of the neutral volatiles
of Emmentaler cheese by the calculation of odor activity values, Lebensm. Wiss.
und Technol. 27:237 (1994).
83. A. Ott, L. B. Fay, and A. Chaintreau, Determination and origin of the aroma impact
compounds of yogurt flavor, J. Agric. Food Chem., 45:850 (1997)
84. G. Reineccius, Off-flavors in foods, in: Source Book of Flavors, 2nd ed. (G. Reinec-
cius, ed.), Chapman & Hall, New York, 1994, pp. 116138.
85. G. Reineccius, Off-flavors in foods, Crit. Rev. Food Sci. Nutr. 29:381 (1991).
86. B. Nijssen, Off-flavors, in: Volatile Compounds in Food and Beverages (H. Maarse,
ed.), Marcel Dekker, New York, 1991, p. 689.
87. Off-Flavors in Foods and Beverages (G. Charalambous, ed.), Elsevier, New York,
1992.
88. R. Marsili, Off-flavors and malodors in foods: mechanisms of formation and analyt-
ical techniques, in: Techniques for Analyzing Food Aroma (R. Marsili, ed.), Marcel
Dekker, New York, NY, 1997, p. 237.
89. S. Widder and W. Grosch, Precursors of 2-nonenals causing the cardboard off-
flavor in butter oil, Nahrung 41:42 (1997).
90. a). A. Stephan and H. Steinhart, Identification of character impact odorants of dif-
ferent soybean lecithins, J. Agric. Food Chem. 47:2854 (1999); b). W. L. Boatright
and Q. Lei, Compounds contributing to the beany odor of aqueous solutions of soy
protein isolates, J. Food Sci. 64:667 (1999).
91. H. Guth and W. Grosch, Deterioration of soya-bean oil: quantification of primary
flavour compounds using a stable isotope dilution assay, Lebensm. Wiss. und Tech-
nol. 23:513 (1990).
92. W. E. Neff and E. Selke, Volatile compounds from the triacylglycerol of cis,cis
9,15-linoleic acid, J. Amer. Oil Chem. Soc. 70:157 (1993).
93. S. Noel, C. Liegeois, G. Lermusieau, E. Bodart, C. Badot, and S. Collin, Release
of deuterated nonenal during beer aging from labeled precursors synthesized in the
boiling kettle, J. Agric. Food Chem. 47:4323 (1999).
94. K. Harayama, F. Hayase, and H. Kato, Contribution to stale flavor of 2-furfuryl
ethyl ether and its formation mechanism in beer, Biosci. Biotech. Biochem. 59:
1144 (1995).
95. K. Jensen, M. A. Petersen, L. Poll, and P. B. Brockhoff, Influence of variety and
growing location on the development of off-flavor in precooked vacuum-packed
potatoes, J. Agric. Food Chem. 47:1145 (1999).
96. C. Masanetz, H. Guth, and W. Grosch, Fishy and hay-like off-flavors of dry spin-
ach, Z. Lebensm. Unters. Forsch. 206:108 (1998).
97. C. Masanetz, H. Guth, and W. Grosch, Hay-like off-flavor of dry parsley, Z.
Lebensm. Unters. Forsch. 206:114 (1998).
98. M. Kamimura and H. Kaneda, Off-flavors in beer, in: Off-Flavors in Foods and
Beverages (G. Charalambous, ed.), Elsevier, New York, 1992, p. 433.
99. S. Masuda, K. Kikuchi, K. Harayama, K. Sakai and M. Ikeda, Determination of
lightstruck character in beer by gas chromatography-mass spectrometry, J. Amer.
Soc. Brew. Chem. 58:152 (2000).
100. J. Templar, K. Arrigan, W. J. Simpson, Formation, measurement and significance
of lightstruck flavor in beer: a review, Brewers Digest 70:18 (1995).
101. D. A. Escudero, P. Hernandez-Orte, J. Cacho, and V. Ferreira, Clues about the role
of methional as character impact odorant of some oxidized wines, J. Agric. Food
Chem. 48:4268 (2000).
102. P. Perpete and S. Collin, Contribution of 3-methylthiopropionaldehyde to the worty
flavor of alcohol-free beers, J. Agric. Food Chem. 47:2374 (1999).
103. H.-D. Belitz and W. Grosch, Food Chemistry, 2nd English ed., Springer-Verlag,
Berlin, 1999, p. 322.
104. R. L. Swaine, T. S. Myers, C. M. Fischer, N. Meyer, and B. Pohlkamp, The forma-
412 McGorrin
tion of off-flavors in fortified juice-containing beverages: identification of 4-

methyl-2-isopropylthiazole, Perfumer & Flavorist, 22:57 (1997).
105. K. Bernath, Sulfurous off-flavours in wine. [Das Boeckser-aroma in Wein], Diss.
Abstr. Intl. C 59:308 (1998).
106. A. Anocibar-Beloqui, P. Guedes de Pinho, and A. Bertrand, Bis(2-hydroxyethyl)
disulfide, a new sulfur compound found in wine. Its influence in wine aroma, Amer.
J. Enol. Viticult. 46:84 (1995).
107. E. Chambers IV, E. C. Smith, L. M. Seitz, and D. B. Sauer, Sensory properties of
musty compounds in food, in: Food Flavors: Formation, Analysis, and Packaging
Influences (E. T. Contis, C.-T. Ho, C. J. Mussinan, T. H. Parliment, F. Shahidi,
and A. M. Spanier, eds.), Elsevier, New York, NY, Developments in Food Science,
Vol. 40, 1998, p. 173.
108. L. H. Aung, J. L. Smilanick, P. V. Vail, P. L. Hartsell, and E. Gomez, Investigations
into the origin of chloroanisoles causing musty off-flavor of raisins, J. Agric. Food
Chem. 44:3294 (1996).
109. J. C. Spadone, G. Takeoka, and R. Liardon, Analytical investigation of Rio off-
flavor in green coffee, J. Agric. Food Chem. 38:226 (1990).
110. S. W. Lloyd and C. C. Grimm, Analysis of 2-methylisoborneol and geosmin in
catfish by microwave distillationsolid phase microextraction, J. Agric. Food
Chem. 47:164 (1999).
111. Y. Jirawan and N. Athapol, Geosmin and off-flavor in Nile tilapia (Oreochromis
niloticus), J. Aquat. Food Prod. Technol. 9:29 (2000).
112. M. R. Sevenants and R. A. Sanders, Anatomy of an off-flavor investigation: The
medicinal cake mix, Analytical Chem. 56:293A (1984).
113. R. J. McGorrin, T. R. Pofahl, and W. R. Croasmun, Identification of the
musty component from an off-odor packaging film, Analytical Chem. 59:1109A
(1987).
114. W. Stark and D. A. Forss, A compound responsible for metallic flavor in dairy
products. I. Isolation and identification, J. Dairy Res. 29:173 (1962).
115. O. W. Parks, D. P. Schwartz, and M. Keeney, Identification of o-aminoaceto-
phenone as a flavor compound in stale dry milk, Nature 202:185 (1964).
116. E. H. Ramshaw and E. A. Dunstone, Volatile compounds associated with the off-
flavor in stored casein, J. Dairy Res. 36:215 (1969).
117. B. Dollmann, A. Schmitt, H. Koehler, and P. Schreier, [Formation of the untypical
ageing off-flavour in wine: generation of 2-aminoacetophenone in model studies
with Saccharomyces cerevisiae], Wein-Wiss. 51:122 (1996).
118. M. Y. Jung, S. H. Yoon, H. O. Lee, and D. B. Min, Singlet oxygen and ascorbic
acid effects on dimethyl disulfide and off-flavor in skim milk exposed to light, J.
Food Sci. 63:408 (1998).
119. T. Koenig, B. Gutsche, M. Hartl, R. Huebscher, P. Schreier, and W. Schwab, 3-
Hydroxy-4,5-dimethyl-2(5H)-furanone (sotolon) causing an off-flavor: elucidation
of its formation pathways during storage of citrus soft drinks, J. Agric. Food Chem.
47:3288 (1999).
120. R. G. Buttery, W. F. Haddon, R. M. Seifert, and J. G. Turnbaugh, Thiamin odor
and bis(2-methyl-3-furyl)disulfide, J. Agric. Food Chem. 32:674 (1984).
121. C. Heiler and P. Schieberle, Model studies on the precursors and formation of the
metallic smelling (E,Z )-2,6-nonadienol during the manufacture and storage of but-
termilk, Int. Dairy J. 7:667 (1997).
122. A. Sensidoni, G. Rondinini, D. Peressini, M. Maifreni, and R. Bortolomeazzi, Pres-
ence of an off-flavor associated with the use of sorbates in cheese and margarine,
Ital. J. Food Sci. 6:237 (1994).
123. H. Shiratsuchi, Y. Yoshimura, M. Shimoda, K. Noda, and Y. Osajima, Objective
evaluation of off-flavor in spray-dried skim milk powder, J. Jap. Soc. Food Sci.
Technol. 43:7 (1996).
124. F. B. Whitfield, N. Jensen, and K. J. Shaw, Role of Yersinia intermedia and Pseu-
domonas putida in the development of a fruity off-flavor in pasteurized milk, J.
Dairy Res. 67:561 (2000).
125. D. A. Zabolotsky, L. F. Chen, J. A. Patterson, J. C. Forrest, H. M. Lin, and A. L.
Grant, Supercritical carbon dioxide extraction of androstenone and skatole from
pork fat, J. Food Sci. 60:1006 (1995).
126. Y. Tanaka, T. Sasao, T. Kirigaya, S. Hosoi, A. Mizuno, T. Kawamura, and H.
Nakazawa, Analysis of skatole in off-flavor beef by GC/MS, J. Food Hygienic Soc.
Jap. 39:281 (1998).
127. A. M. Popken, H. M. Dechent, and D. Guerster, Investigations on the origin of
carvone in orange juices as an off-flavor component, Fruit Processing 9:338 (1999).
128. M. Walsh, R. Rouseff, and M. Naim, Determination of furaneol and p-vinylguaiacol
in orange juice employing differential UV wavelength and fluorescence detection
with a unified solid phase extraction, J. Agric. Food Chem. 45:1320 (1997).
129. J. A. Donaghy, P. F. Kelly, and A. McKay, Conversion of ferulic acid to 4-vinyl
guaiacol by yeasts isolated from unpasteurised apple juice, J. Sci. Food Agric. 79:
453 (1999).
130. D. Madigan, I. McMurrough, and M. R. Smyth, Rapid determination of 4-vinyl
guaiacol and ferulic acid in beers and worts by high-performance liquid chromatog-
raphy, J. Amer. Soc. Brew. Chem. 52:152 (1994).
Abbreviations
ACT automated calibration transfer

AEDA aroma extraction dilution analysis, a GC-O technique
2-AP 2-acetyl-1-pyrroline
APC all-purpose cleaner
ASE accelerated solvent extraction
CAR Carboxen SPME fiber
CHARM CharmAnalysis, Combined Hedonic Aroma Response
Measurements, a GC-O technique
CI chemical ionization (mass spectrometry)
CLSA closed-loop stripping analysis
CW Carbowax
2D-GC two dimensional GC
DCM dichloromethane
DH dynamic headspace
DTD direct thermal desorption
DVB divinylbenzene
ECD electron capture (GC) detector
ECID external closed-loop inlet device
EDT eau-de-toilette
EI electron impact (mass spectrometry)
E-nose electronic nose
FID flame ionization detector
FD flavor dilution (factor), a GC-O technique
FPD flame photometric (GC) detector
415
416 Abbreviations
GC gas chromatography
GC-MS GC-mass spectrometry
GC-O GC-olfactometry
GC-TOFMS GC with time-of-flight mass spectrometry detection
HCA hierarchical cluster analysis, an MVA technique
HRGC high resolution GC
HSSE headspace sorptive extraction
IDA isotope dilution analysis
ITD ion trap (mass spectrometry) detector
ITMS ion trap mass spectrometer
ITR integrating transient recorder
KNN K nearest neighbor, an MVA technique
LSD least significant difference
MS mass spectrometry
MSD mass spectrometry detector
MVA multivariate analysis
m/z mass to charge ratio
NIF nasal impact frequency, a GC-O technique
NPD nitrogen phosphorous (GC) detector
OAV odor activity value, a GC-O technique
Osme from the Greek word , meaning smell, a GC-O
technique
PA polyacrylate
PCA principal component analysis, an MVA technique
PCI positive chemical ionization (MS)
PCR principal component regression, an MVA technique
PDMS polydimethylsiloxane
PLS partial least squares, an MVA technique
P&T purge-and-trap
R&G roasted and ground (coffee)
RSD relative standard deviation
RTIC reconstructed total ion chromatogram
SAFE solvent-assisted flavor evaporation
SBSE stir bar sorptive extraction
SCD sulfur chemiluminescence (GC) detector
SFE supercritical fluid extraction
SDE simultaneous steam distillation/extraction
SH static headspace (GC)
SIDA stable isotope dilution analysis
SIM selective ion monitoring, a GC-MS technique
SIMCA soft independent modeling of class analogy, an MVA
technique
Abbreviations 417
SNIF surface of nasal impact frequency, a GC-O technique

SLES sodium lauryl ethoxy sulfate
SPE solid phase extraction
SPME solid-phase microextraction
SPME-MS-MVA solid-phase microextraction, mass spectrometry,
multivariate analysis
SPTD short-path thermal desorber
TAD time-array detection
TIC total ion chromatogram
TMP 2,4,6-trimethylpyridine
TCC time-compressed chromatography
TOF time-of-flight (MS)
TOFMS (ToF MS) time-of-flight MS
TMP 2,4,6-trimethylpyridine
TSD time-slice detection
VFA volatile fatty acids
WOF warmed-over flavor
Index
Accelerated solvent extraction (ASE), Beer, off-flavors, 399

140 Bell pepper, 385386
2-Acetyl-1-pyrroline in rice (2-AP), Benzothiazole, 281
238, 245246, 390 Blackcurrant, 385
Allium, 317, 380381 Black peppercorn, 102103
Amersorb, 36 Bread, 305307
Anethole, 81 Bread crust, 391
Apple, 188189 Breakthrough volume, 36
delicious, 381382 BTEX compounds, 85
Apricot flavor, by static headspace and
SPME, 146148
Aroma extraction dilution analysis Cabbage, 387
(AEDA), 140, 299, 301302 Canary melon, 9294
Aromagram, 334, 337, 340 Cantaloupe, 9899
Artifacts, 75 Carbopack, 36
Automated calibration transfer (ACT), Carbosieves,36
361 Carbotrap, 36
Carbowax, 78, 142
Carboxen, 36, 142, 213219, 369
Banana: Carboxylic acids, volatiles, by SPME,
dynamic headspace of, 46, 48 221
SPME analysis of, 98100 Carrot, 387
Basil, 378 Cat ketone, 375
Beef, grilled ground, by direct thermal Cauliflower, 387
desorption, 64, 6768 (see also Celery, 387388
Warmed-over flavor) Champagne, 334335, 338, 341, 343
419
420 Index
Character impact compounds, 375405 Dairy products (see also Milk):

CharmAnalysis, 140 impact flavor compounds in, 395398
Cheese, 395398 off-flavors in, 404405
Cheddar, 376 -Decalactone, 81
dynamic headspace of, 4344 Dendrogram, 354
by SAFE, 164166 Deodorants, 289291
with SIDA for quantitation, 187 Direct injection of sample, 45
Cherries, 385 compared to SPME headspace sam-
Chile powder, 378 pling, 81
Chinese mountain pepper, by SPME, compared to SPME liquid sampling,
146147 81
Chive, 380381 Direct thermal desorption (DTD), 55
Chocolate, 376, 389390, 392 71
Cleaner, all-purpose, 287288 injectors for, 6364
Closed-loop stripping analysis (CLSA), quantitative analysis, 68, 71
141 repeatability, 69
of flowers, 253254, 256 sample constraints, 60
Coffee: short-path thermal desorber (SPTD),
aroma impact odorants of roast and 59
ground Arabica and Robusta, substances analyzed by, 61
303304 Distillation:
character impact compounds in, 389 co-distillation of sample with solvent,
390 2223
classes of aroma compounds in, 2 short path distillation, 2022
classification of ground coffee sam- simultaneous steam distillation/extrac-
ples with SIMCA, 356 tion (SDE), 1418
espresso ground coffee, 206207 under reduced pressure, 312
green, 338 versus static headspace, 314
green and roasted, comparison of, 16 with direct solvent extraction, 312
by SPME using SIDA for quantita- 314
tion, 187, 190 steam distillation followed by solvent
by SPME using three different fibers, extraction, 813, 19
220221 vacuum steam distillation, 910
total ion chromatogram of, 6 Divinylbenzene (DVB), 78, 142
Cola, 103105 Dynamic headspace (see Headspace sam-
Combi PAL, 278, 371372 pling techniques)
Competition phenomena, 222223
Comprehensive 2-D chromatography, Eau de toilette, by SPME, 147149
115 Electronic nose (e-nose) with MS sen-
with TOFMS, 132133 sor, 349373
Consumer products, 277293 advantages of, 373
Cryogenic trapping, 3740 automated calibration transfer (ACT),
Cucumber, 386 361
Curry powder, 103104 chemical ionization (CI) techniques,
Cyclotene, 317, 319 356360
Index 421
[Electronic nose (e-nose) with MS sensor] [Fruit]

shelf-life prediction of milk, 364371 fresh, 97102
SPME-MS-MVA, 361374 juice, 206, 208
Enfleurage, 249250 Furaneol, 308309, 315, 320
Epiphyllium oxpetalum, 272
Essential oils, of flowers, 249 Garlic, 380381
Ethanol, by SPME, 8789 Gas chromatography-olfactometry
Ethyl vanillin, 81, 83 (GC-O), 297325, 333374
EZ-Flash, 166168 aroma extraction dilution analysis
(AEDA), 301302, 333335
Fabric softener, 189, 191192 CharmAnalysis, 299300, 333, 335
Fast chromatography, 129130, 168, 336
371373 chromatographic problems, 344345
Fast-MS, 371373 dry fruit flavor, by SPME, 155
Fenugreek, 378 effect of chromatography, 316319
Flavor dilution (FD) factor, 301302, interpretation of, 323324
317319, 334 limitations of, 322323
coffee aroma impact odorants, 304 Osme, 298299, 334, 336
Flowers, 249276 pickles, by SPME, 214, 216219
Asian lily, 263 quantitative ability of AEDA, Charm,
biorhythm, 264265 Osme, 334336
hyacinth (blue) SNIF method, 333374
comparison of Zenith trap and head- advantages, 336337, 345
space, 261 comparison with other methods,
by SPME, 258259 343
lavender, by SPME, 152154 principle, 337338
lily, by SPME, 152154, 263 quantitative olfactogram compari-
living versus cut flowers, 261 sons, 338341
maturity of the flower, 263264 screening impact compound with,
rose, by SPME, 152154 338
rose, hybrid tea, 263 techniques, 298304
scent collection in rain forest cano- GC-TOFMS, 107138
pies, 265270 applications of, 129136
scent, reconstitution, 270273 data analysis, 119129
scents by SPME, 258260 determination of individual-compo-
Zenith trap, 260261 nent elution profile, 121122
Formaldehyde, 282284 determination of the correct frag-
Fragrance: mentation pattern, 122125
stability, 278279 peak finding, 120121
substantivity qualitative analysis, 125128
cloth/tile, 288289 quantitative analysis, 128129
skin, 289291 data collection, 119
Fruit: deconvolution of peaks, 125126,
character impact chemicals in, 382 133136
flavor, 381385 minimum acquisition rates with, 114
422 Index
[GC-TOFMS] -Ionone, 81
principles of, 116119 Isoprene, 251252
techniques of GC-MS, 113115 Isothiocyanates, in seafood cocktail
types of mass spectrometers, 111 sauce, by SPME, 8990
113 Isotope dilution analysis (IDA), 180
Geosmin, 55 (see also Stable isotope dilution
Grape, 178, 381382 analysis)
Grapefruit juice, 166168, 173, 382,
384 Jasmine, 250
Guaiacol, in coffee, 190
Kiwi:
Headspace sampling techniques, 2552
by dynamic headspace, 48
dynamic, 2952
by SFE, 161
analysis of flowers, 252258
K nearest neighbor (KNN), 355
advantages of, 30
Kutscher-Steudel, rotational performa-
breakthrough volume, 36
tion, 140
disadvantages of, 3031
liquid samples, 3132
Lactones, 384, 387, 393, 395396
solid samples, 3234
Leek:
water management, 38
compounds in, 380381
with SPME, 291293
by SFE, 160161
purge-and-trap:
Lemon:
selection of trapping technique, 34
flavor-impact compounds in, 384
sorbent trapping, 3436
volatiles collected from, 49
quantitative analysis, 5152
Likens-Nickerson, 16, 23, 140
static, 2629
modified, 19
analysis of dry apricot flavor, com-
of punch flavor, 97
pared to SPME, 146148
analysis of flowers, 252
Maillard flavors, 389392
advantages of, 28
Mass spectrometers:
disadvantages of, 2829
as an e-nose sensor, 351374
Headspace sorptive extraction (HSSE),
types of, 111113
76
Meat, 362364, 393395
Heliotropin, 81
1-p-Menthene-8-thiol in grapefruit juice,
Herbs, 377381
305, 382, 384
-Hexalactone, 81
4-Mercapto-4-methyl-2-pentanone, 375
Hierarchical cluster analysis (HCA),
Methionyl acetate, 9294
353, 372373
Methylisoborneol, 55
High-pressure extraction with supercriti-
Microwave extraction, 140
cal CO2 (see Supercritical fluid ex-
Milk, 213215 (see also Dairy prod-
traction)
ucts)
Homofuraneol, 308309
aromagram, 340
Horseradish, 8990
shelf-life prediction by SPME-MS-
Household cleaning products, 147151
MVA, 364371
Indole, in dehydrated yeast, by SPME, Mixxor, 1011
9092 Monterpene, formation in plants, 252
Index 423
Multichannel detection, 108113 Polydimethylsiloxane (PDMS), 78, 142

Multivariate analysis, 367, 372 Popcorn, 389390
Musty aromas and flavors, 403404 Porapak Q, 255257, 273
Positive chemical ionization (PCI),
Nafion tubing, 40 356360
Nasal impact frequency (NIF), 337 Potato, 387
Nootkatone, 384 n-Propanol, by SPME, 8789
Nuts, 392393 Principal component analysis (PCA),
353355, 373
Ocimene, 256 warmed-over flavor in beef, 362364
Octatrienols, 255 Principal component regression (PCR),
1-Octen-3-one, 342 354355
Odor activity value (OAV), 335 Psychrotrophic bacteria, 370
Odor thresholds, of some aroma impact Punch flavor, 9597
components determined by GC-O, Purge-and-trap (see Headspace sampling
321 techniques)
Off-flavor/off-odor analysis, 278283, -Pyrrolidone in rice, 231
310311, 349351, 398405
Olfactogram, 333334, 337339 Raspberry, 384
Olfactometry (see Gas chromatography Recombination studies, 219
olfactometry) Rice, 229246
Onion, 380381 brown rice, 242243
Oral malodor, 284286 compounds in the headspace of, 239
Osme, 140, 298299 242
milled rice, 242243
Packaging malodor, 281282, 349 sample preparation, 234235
Parsley, 388 SPME optimization, 235236
Partial least squares (PLS), 354 waxy rice, 243244
shelf-life prediction of milk, 364371 Rose:
Partition coefficient, 76 living flower, by SPME, 152154
Passion fruit flavor compounds: oil, by SPME, 15355
flavor compounds in, 384385 petals, by SFE, 160163
by SPME, 146
Peanuts, by direct thermal absorption, SAFE (see Solvent-assisted flavor evapo-
6667 ration)
Pear, Bartlett, 99100 Saffron, 378
Peas, 386 Salting-out effect, 8384, 143
Pepper, 378 Saliva, 284286
Peppermint oils, 358359 Sample preparation, complicating fac-
Pickles, 214, 216219 tors, 12
Pineapple, 382 Seafood, 393395
Polyacrylate (PA), 142 Seafood cocktail sauce, isothiocyanates
comparison of results with PDMS fi- by SPME, 8990
ber, 80 Seasoning, 377381
for carboxylic acids, 83 Scalping, 350
for polar compounds, 80 Short-path thermal desorber (SPTD), 59
424 Index
Simultaneous steam distillation/extrac- [Solid-phase microextraction (SPME)]

tion (SDE) (see Distillation) headspace vs. immersion for carbox-
Skatole, in coffee, 190, 193 ylic acids, 8384, 141
SNIF method (see Gas chromatography- information on the World Wide Web,
olfactometry) 76
Soap, fragrance chemicals in, 147151 of juice, 206, 208
Sodium lauryl ethoxy sulfate (SLES), manual method, 232233
279280 of milk, 213215, 366
Soft independent modeling of class anal- parameters to optimize, 141145
ogy (SIMCA), 355356 pH, effect on sample extraction, 143
peppermint oils, 358359 of pickles, 214, 216219
Solid-phase extraction (SPE), 140, 214, qualitative analysis, 94105, 234
216219 244
Solid-phase microextraction (SPME), quantitative analysis, 8794, 144,
75106, 141156, 277293 244246
automated method, 233234 salting-out effect, 8384, 143
carboxylic acids, extraction effective- sampling time, 79
ness with different fibers, 8284, theoretical considerations, 8486
221 of tomatoes, 209212
of coffee, 206 types of fibers available, 142
using SIDA for quantitation, 187, of vegetable oil, 207209
190 what it is, 76
comparison studies Solvent-assisted flavor evaporation
with different fibers, 205226 (SAFE), 163
with other sample preparation meth- beverages, 166
ods, 205226 cheese, 164
competition phenomena, 222223 versus SDE, 169171
of consumer products, 277293 washing powder, 169
contaminants, 236238 Solvent extraction:
cryogenic cooling with, 80 common solvents, 5
derivatization directly on fiber, 282 concentration, 8
284 direct extraction of aqueous samples,
of dry aprocit flavor, compared to 58
static headspace, 146148 emulsions, 8
extraction steps and desorption, 78 impurities, 8
84 liquid carbon dioxide as solvent,
extraction temperature and time, im- 68
portance of controlling, 88, 143 liquid-liquid extraction, 210, 212
144 of solid samples, 1819
fiber selection: Sorbent trapping, 3436
based on analyte characteristics, Sotolone, 378, 389
223226 Spices, 377381
guide, 226 SPME-MS-MVA (see Electronic nose
of flowers, 258260 with MS sensor)
graphical representation of a SPME Stable isotope dilution analysis (SIDA),
device, 77 9294, 179194
Index 425
Static headspace sampling (see Head- Trace analyses, 130132

space sampling techniques) Trichloroanisole, 403
Steam distillation (see Distillation) 2,4,6-Trimethyl pyridine (TMP), 243,
Stir bar sorptive extraction (SBSE), 76, 246
171179 Truffles, 209210
Strawberry, 381382 Turmeric, 378
Strecker degradation, 309, 399, 403 Two-dimensional (2D) gas chromatogra-
Styrofoam cup/packaging, dynamic phy, 114115
headspace of, 5051
Sugar, beet, 6466 -Undecalactone, 342
Sulfur compounds, in saliva, 285286
Supercritical fluid extraction (SFE),
157163 Vacuum distillation (see Distillation)
Vanillin, 81, 83
Tedlar plastic bags, 266270 Vegetables, character impact flavor com-
Tenax, 35, 255260 pounds in, 385388
Thai jasmine rice, 238 Volatile fatty acids (VFAs), 369
Thyme, 378
Time-array detection (TAD), 117119 Warmed-over flavor (WOF), 310311,
Time-compressed chromatography 362364, 398
(TCC), 114 Wine:
Time of flight mass spectrometer (see Cabernet Sauvignon, 385
GC-TOFMS) classification according to vineyard,
Time-slice detection (TSD), 117119 355357
TOF (see GC-TOFMS) Wintergreen, 377
Tomatillo, 386387
Tomatoes, 209212 Yogurt, aromagram, 340
Toothpaste, 156
dynamic headspace analysis of, 43 Zenith Trap, 141, 260262, 270

Flavor, Fragrance, and Odor Analysis

Uploaded by

Copyright:

Available Formats

Flavor, Fragrance, and Odor Analysis

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Flavor, Fragrance, and Odor Analysis

Uploaded by

Copyright:

Available Formats

ISBN: 0-8247-0627-7

This book is printed on acid-free paper.

Eastern Hemisphere Distribution

World Wide Web

Copyright 2002 by Marcel Dekker, Inc. All Rights Reserved.

Current printing (last digit):

PRINTED IN THE UNITED STATES OF AMERICA

For various reasons, reliable electronic-nose applications have been slower

1. Solvent Extraction and Distillation Techniques 1

2. Analysis of Food Volatiles Using Headspace-Gas

3. The Analysis of Food Volatiles Using Direct Thermal

4. Solid-Phase Microextraction for the Analysis of Aromas and

5. The Advantages of GC-TOFMS for Flavor and Fragrance

6. Modern Methods for Isolating and Quantifying Volatile Flavor

7. SPME Comparison Studies and What They Reveal 205

8. Analysis of Volatile Compounds in the Headspace of Rice

9. Headspace Techniques for the Reconstitution of Flower Scents

10. SPME Applications in Consumer Products 277

11. Gas ChromatographyOlfactometry in Food Aroma Analysis 297

12. Quantitative Use of Gas ChromatographyOlfactometry: The

13. Combining Mass Spectrometry and Multivariate Analysis to

14. Character Impact Compounds: Flavors and Off-Flavors in Foods 375

Heinz-Jurgen Bertram Haarmann & Reimer GmbH, Holzminden, Germany

Imre Blank Nestle Research Center, Lausanne, Switzerland

Stefan Brennecke Haarmann & Reimer GmbH, Holzminden, Germany

Wilfried Bretschneider Haarmann & Reimer GmbH, Holzminden, Germany

Alain Chaintreau Firmenich S.A., Geneva, Switzerland

Elaine T. Champagne Southern Regional Research Center, Agricultural Re-

Ben D. Gardner Orion Associates Science & Engineering, Okemos, Michigan

Casey C. Grimm Southern Regional Research Center, Agricultural Research

John F. Holland Department of Biochemistry, Michigan State University, East

John Labows Colgate-Palmolive Company, Piscataway, New Jersey

Steven W. Lloyd Southern Regional Research Center, Agricultural Research

Ray Marsili Dean Foods Company, Rockford, Illinois

Thomas McGee Global Technology and Innovation, Givaudan Fragrances

Robert J. McGorrin Department of Food Science and Technology, Oregon

James A. Miller Southern Regional Research Center, Agricultural Research

Kenichi Ohtsubo National Food Research Institute, Ministry of Agriculture,

Thomas Parliment Parliment Consulting, New City, New York

Richard Payne Colgate-Palmolive Company, Piscataway, New Jersey

Allen E. Puchalski Colgate-Palmolive Company, Piscataway, New Jersey

Kenneth L. Purzycki Givaudan Fragrances Corp., Teaneck, New Jersey

Arthur M. Spanier Beltsville Agricultural Research Center, Agricultural Re-

Thomas P. Wampler CDS Analytical, Inc., Oxford, Pennsylvania

Peter Werkhoff Haarmann & Reimer GmbH, Holzminden, Germany

TABLE 1 Classes of Aroma Compounds in Coffee

This is particularly important since current flavor research seems to be less

II. DIRECT INJECTION OF THE SAMPLE

III. DIRECT SOLVENT EXTRACTION OF AQUEOUS

FIGURE 3 Liquid/liquid extractor concentrator apparatus. (Courtesy Supelco, Inc.,

IV. STEAM DISTILLATION OF SAMPLES FOLLOWED

One of the most common sample-preparation techniques employed today in-

B. Indirect Steam Distillation

C. Vacuum Steam Distillation

is normally not a problem in this case. The higher-boiling components do not

E. Manipulation of the Aqueous Phase

FIGURE 5 Comparison of chromatograms of etheral (upper) and methylene chloride

FIGURE 6 Comparison of chromatograms of R&G coffee extracted at pH 3.6 (upper

V. SIMULTANEOUS STEAM DISTILLATION/EXTRACTION